You are on page 1of 35

Composites Science and Technology 64 (2004) 343377

The predictive capability of failure mode concept-based strength

criteria for multidirectional laminates
R.G. Cuntze*, A. Freund
Main Department Analysis, MAN Technologie AG, Franz Josef Strau Str. 5, D-86153 Augsburg, Germany

Received 1 October 1998; accepted 1 November 2001

This contribution is a post-runner to the failure exercise. It focuses on two aspects of the theoretical prediction of failure in
composites [13]: the rst is the derivation of failure conditions for a unidirectional (UD) lamina with the prediction of initial fail-
ure of the embedded lamina, and the second the treatment of non-linear, progressive failure of 3-dimensionally stressed laminates
until nal failure. The failure conditions are based on the so-called Failure Mode Concept (FMC) which takes into account the
material-symmetries (by the application of invariants) of the UD-lamina homogenized to a material, and on a strict failure mode
thinking. The results of the investigation are stressstrain curves for the various given GFRP-/CFRP-UD-laminae, biaxial failure
stress envelopes for the UD-laminae, and initial as well as nal biaxial failure envelopes for the laminates. In addition a brief
comparison between Pucks and Cuntzes failure theory is presented by the authors themselves.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: B. Non-linear behaviour; Multi-axial stressing; Multidirectional laminates; C. Failure criterion

1. Introduction mality criterion: the subsequent failure surface is indi-

cated by a vector normal to the actual global yield
For a reliable Strength Proof of Design of a laminate failure surface) is replaced by the idea of proportional
composed of UD-laminae reliable failure criteria and a stressing, that means, the surface increases in the direc-
reliable progressive failure analysis are needed. tion of the actual stressing which is seldom the normal
The non-linear behaviour of laminates composed of direction. Partial mode-related fracture surfaces will
brittle laminae (these are the materials in the failure conne the subsequent global anisotropic yield surface
exercise) originates from the damage development piecewise. These fracture surfaces are essentially descri-
around inherent defects in the constituent matrix and at bed by those fracture conditions for the UD lamina
the interface bre-matrix (ductile matrix materials (dened here to be the material the laminate consists of)
would show necking and so-called crazing, which which are matrix-dominated.
appears in case of glass bre composites as whitening in The development of UD failure criteria and of degra-
a tensile test). These defects grow to micro-cracks and dation models for the progressive failure analysis gave
later to cracks under increased stressing. Therefore, the rise to activities in Germany. These activities con-
usually in the plasticity theory to be applied global yield centrated as far as possible on the improvement of fail-
failure condition [14] which has to be anisotropic here ure criteria and their verication by multi-axial testing
(see Hills yield condition with associated ow rule), is with the existing specimens and test rigs [511].
to be replaced by a single global or several partial fracture Since 1980 a group close to Puck has tried to improve
conditions. Also, the so-called associated ow rule (Nor- Pucks old criteria [8,9] which already distinguished
between the two failure types, bre failure (FF) and
inter bre failure (IFF). In 1992, Puck [12] eventually
* Corresponding author. Tel.: +49-821-505-2593; fax: +49-821-
established his new set of IFF-criteria following an
505-2630. idea, proposed 1980 by Hashin [13], which was based on
E-mail addresses: (R.G. Cuntze). a modied Mohr/Coulomb theory.
0266-3538/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
344 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

Nomenclature vf volume fraction

In the nomenclature, self-explaining symbols are used x1, x2, x3 Coordinate system of a unidirectional
if a property is addressed. A lamina (is dened to be (UD-)lamina (x1=bre direction,
the calculation unit) may consist of several physical x2=direction transverse to the bre,
layers. x3=thickness direction)
"1, "2, "3 Normal strains of a unidirectional
Unidirectional lamina lamina
as ; bs Ramberg/Osgood parameters in 12 Major Poissons ratio in the failure
softening regime exercise (corresponds to ?k in the
b? ; b?k ; b?k Curve parameters German guideline VDI 2014. There
E1 Ek , Elastic moduli of a UD lamina in is no rationale for 12 or 21 . In the
E2 E3 E? the directions x1, x2, x3 early times the application of 21
E1tan ; E3sec A tangent and a secant elastic was preferred because this
modulus denotation makes more sense
Eff res Resultant stress eort of all (location rst, cause second)
interacting failure modes.  1,  2,  3 Normal stresses in a unidirectional
Corresponds to Pucks exposure layer
factor fE 2c ; 1t Compressive stress across, tensile
Eff mode Stress eort of a UD-lamina in a stress in bre direction
failure mode, e.g. eq =Rkc Eff k . k ; ? Stresses parallel and transverse to
corresponds to 1=f kif linear the bre direction
behaviour ^; ^ Laminate mean stresses
maxEFFmode Stress eort of the maximum f  g L ; f  g R Load-dependent stresses; residual
stressed failure mode stresses
etk ; eck tensile and compressive failure eq Equivalent stresses of a mode
k k ? ? ?k
strain of a UD-lamina in x1 eq ; eq ; eq ; eq ; eq ; includes
direction load induced mechanical stresses and
Fk , Fk , F? , Failure functions for FF and IFF residual stresses
F? , F?k 12 21 , Shear stresses of a unidirectional
13 31 , lamina in the elastic symmetry
f Mode
Res Reserve factor=stretching factor for 23 32 directions. The rst subscript locates
the applied stress state necessary to the direction normal to the plane on
achieve the failure stress state of the which the shear stress is acting; the
t ?
mode, e.g. f?
Res R? =eq 1 second subscript indicates the
fE Stress exposure factor of Puck direction of the shear force
f Res Resultant reserve factor of all ?k ; ?? Shear stressing transverse/parallel
interacting failure modes and transverse/transverse to the
G21 ; G21sec Shear modulus of a UD lamina in bre direction
the x2, x1 direction, secant shear 12 21 , Shear strains of a unidirectional
modulus 13 31 , layer
I1 ; I2 ; I3 ; I4 ; I5 Invariants of the transversally- 23 32
isotropic UD-material
MS Margin of safety fRes  1 Characteristics of the bres
m Mode interaction coecient E1f Elastic modulus in x1 direction
Rp0:2 Stress value at 0.2% plastic strain 1f ; 2f Stress in x1 direction; stress in x2
Rkt  X t , UD tensile and compressive (basic) direction.
Rkc  X c strength parallel to the bre
R?t  Y t , UD tensile and compressive strength Potential fracture plane (for the comparison PuckCuntze)
R?c  Y c transverse to the bre direction R?A ; R?A Fracture resistance of the action
R?k  S Shear strength of a UD lamina plane against its fracture due to
transverse/parallel to the bre transverse tensile and compression
direction stressing. They correspond to
R; R Mean strength, design allowable strength values R?t , R?c .
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 345

R?k Fracture resistance of the action FF Fibre failure
plane against its fracture due to FMC Failure mode concept
transverse/parallel shear stressing FoS Factor of safety
: R?k FPF First ply failure
R?? Fracture resistance of the action FRP Fibre-reinforced plastic
plane against its fracture due to IFF Inter-bre failure
transverse/transverse shear stressing MS Margin of safety
x1, xn, xt Coordinate system rotated with
respect to the bre direction by an
angle from the x2 direction to the Indices, signs
xn direction A indicates an Action plane quantity
n ; nt ; nl Normal stress, normal/longitudinal c, t compression, tension (German
shear stress, normal/transverse shear Guideline VDI 2014)
stress acting on the potential f, m bre, matrix
fracture plane (MohrCoulomb fp bre-parallel fracture plane
stresses) (sec) secant modulus
Angle between the x2 axis and the xn (res) resultant
axis s symmetric lay-up, softening
fp Angle of the fracture plane. Res Reserve
^ laminate mean stress or average
stress of laminate
Abbreviations (+), () mathematical notations for tension
CLT Classical laminate theory and compression
CoV Coecient of variation  statistical mean
DLL Design limit load ,  indicate the failure induced by the
F Failure function normal or shear Mohr stress
FEA Finite element analysis

From 1992 [8] to 1997, Cuntze [14] as well as others ring in the isotropic case. Such a global criterion has on
[1522] focussed on the Puck/Hashin IFF-Strength Cri- the one hand numerical advantages because one has to
teria, which are based on the determination of the frac- apply only one criterion, but on the other hand, it may
ture plane. The result of this work was incorporated in lead to erroneous results due to its physical shortcoming
the nal report of a research project [6]. In parallel, because it tries to map several failure modes).
since 1994 Cuntze has investigated invariant-based for- Invariant-based failure criteria have been formulated
mulations of strength criteria for isotropic and aniso- for a large number of isotropic materials. As the rst
tropic materials [2329]. Cuntzes main idea is not the Hashin [13] seems to have postulated (1980) in the same
basement on invariants but the strict allocation of a paper, in parallel to his MohrCoulomb model-based
strength criterion to one failure mode and to one asso- IFF criteria, invariant-based UD-failure criteria. Based
ciated basic strength. on curve tting consideration and not on physical rea-
While Cuntze was studying the invariant-based v. soning, Hashin chose a quadratic approximation which
Mises yield criterionit describes one (strength) failure reads in its general form
mode, the shear yielding, and allows for the determi-
A1 I1 B1 I12 A2 I2 B2 I22 C12 I1 I2 A3 I3
nation of the slip line anglesthe question raised to 0
him: Why should it not be possible to formulate for A4 I4 1;
each single failure mode of an anisotropic material an
appropriate invariant-based mode failure criterion which and which includes six strengths (for the denition of
might probably (a further condition has to be applied) the invariants, see [13]). From the 3D failure criterion
later allow for post-determination of the failure angle, if above he modelled four distinct failure modes: the ten-
desired? sile and compressive bre modes and two matrix modes.
The application of invariants is almost standard for This results in piecewise smooth failure surfaces which
isotropic materials. However there, the main intention is do not t well (Fig. 3 in [13]). The comparison of
to build up a yield criterion (this means for one single Cuntzes results [29] with Hashins formulations show
failure mode or one phenomenon) or a global fracture some dierences: (1) Hashins choice of a single quad-
criterion that includes all fracture failure modes occur- ratic approximation, (2) two matrix modes, (3) six
346 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

strengths (Hashin in reality uses R?? not Pucks R?? . Confronted with various questions of the UD failure
R?? is identical to the strength R? in our actual case of criteria community, Cuntze in cooperation with A. Puck
brittle behaviour), and (4) the application of tensile tries to outline in this contribution (see Appendix A) the
stress  1 combined with longitudinal shear stress  21 coincidences and main dierences of their IFF theories.
(not just the bre tensile stress alone as with Cuntze or Pucks approach usesas proposed by Hashina
Puck). Only for bre parallel compression failure modied Mohr/Coulomb [39] theory for brittle IFF of
Cuntze considers such a contribution of the longitudinal unidirectional (transversally-isotropic) laminae. For
shear stress. But, due to insucient data for this com- IFF thereby is an automatic interaction of stresses
pression FF also the simple maximum stress criterion is included due to basing IFF just on the three so-called
proposed by Cuntze, like Hashin. action plane stresses ( n,  nt,  n1). These stresses have a
Since the early 1980s Boehler et al. [30,31] eventually common action plane (Fig. A1). Therefore, these criteria
extensively pursued the idea of applying invariant-based are called action plane strength criteria. Puck dis-
criteria which they had partly veried by test. Because criminates two fundamental regimes:  n > 0 and  n < 0.
this working group did not present their failure criteria The unknown IFF fracture angle is determined when
in the conventional UD lamina stresses, their valuable the action plane of maximum stress eort is found.
results unfortunately did not attract the stress man. The well-known conventional global criteria apply all
Of course, also Tsai/Wus polynomial failure condi- six stresses of the UD lamina and do not take into con-
tions may be transformed into formulations of invariant sideration whether they might act on the same or on
terms [32]. dierent action planes.
Later in 1996 Cuntze inuenced Jeltsch-Fricker and It is very simple in the plasticity theory of isotropic
Meckbach from the University of Kassel to pick up the materials to develop a so-called single yield failure
idea of invariant-based formulations. They approxi- surface criterion, that means one global criterion, due
mated the Puck/Hashin IFF body by means of two to the existence of only one failure phenomenon, the
invariant formulations[33]. isotropic yielding.
The idea of thinking in strength failure modes is not a A global criterion for fracture may include more than
new idea, but the so-called Failure Mode Concept one fracture failure mode potentially occurring under
(FMC) more strictly applies the mode thinking and the various stress states. It is sometimes also used
more consequently uses the advantage of formulating instead of a global yield criterion in-spite of the facts
the failure conditions (interaction of stresses within a that a fracture criterion as a mathematical description
mode) by the material symmetries respecting invariants, of the fracture failure surface just confines a global
which contain the lamina stresses of the FEM output. yield failurebody (yield capacity exhausted), and that it
This approach, according to the number of the material usually generates a dierent failure shape.
symmetries, requires two independent FF modes and And for laminae? For them, as already mentioned,
three IFF modes. The application to UD material is the instead of a matrix-determined anisotropic global yield
most intensive application of Cuntzes FMC, which is criterion a set of fracture criteria on lamina level is
claimed to be applicable to any material [23,26,27]. applied. These show due to their various failure modes a
Each of Cuntzes ve (see also Christensen [34]) failure multi-fold non-linearity which requires a much higher
modes is characterized by one strength and one eort. A further shortcoming is: A set of failure criteria
modulus. instead of one global one prevents from a simple imple-
The choice of the invariants in Cuntzes FMC is sup- mentation into a commercial FEM code in order to take
ported by physical considerations based upon Beltrami. advantage of the codes solution architecture and pre-/
The decision for an individual basic invariant is directed postprocessor capabilities. This point waits to be tackled.
by the fact whether the material element is subjected in Progressive failure analysis [5860] of laminates or the
the envisaged failure mode to a volume change or a prediction of laminate behaviour up to fracture is the
shape change. major challenge compared to the derivation of reliable
Cuntzes previous experience with structural relia- UD-failure criteria. Cuntze assumes a so-called eective
bility [19,3538], where failure mode thinking is a basic stressstrain curve for the lamina which respects the
idea, helped to simply model the interaction of modes inuence of being embedded [29,40] in the laminate.
within a lamina by the application of a spring model. To be utilized in the non-linear analysis is the secant
Cuntze tries to formulate easy-to-handle homo- modulus which alters for a non-linear stressstrain
geneous invariant-based criteria with stress terms of the curve. However, data are not only needed for the pure
lowest possible order and which make a search of the failure mode domains but for the mode interaction
fracture plane not necessary. The FF criteria are treated domains, too, where the actual stress state aects more
as decoupled from the IFF ones. The interaction of FF than one mode. The inuence of the stress state in a
with IFF is considered probabilistically as within the mode interaction domain on the secant modulus of each
IFF modes by the spring model mentioned above. aected mode is considered by a triggering approach.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 347

This approach increases the equivalent stress (which 2. Main features of the failure mode concept (FMC)
considers all inuencing stresses) of the aected mode in
the case of hardening (the secant modulus becomes a little The features of the FMC are briey summarised in
smaller) and decreases the equivalent stress in the case of Table 1. Additional aspects are collected in Table 2.
softening (the secant modulus becomes smaller, too). The These features and some further aspects will be descri-
modes equivalent stressstrain curve is identical to the bed in the coming sections in more detail. The FMC is a
associated uni-axial stressstrain curve measured. general concept, UD material is one application only.
A crucial dierence between Pucks [42] approach and
Cuntzes approach is the treatment of degradation in
the non-linear-analysis of the laminate. Both theories 3. Basics
apply the self-correcting secant modulus method, how-
ever, describe the successive degradation (the softening) 3.1. State of stress
dierently as well as the rounding-o in the interaction
domains of FF and IFF modes. For the unidirectional (UD) material element Fig. 1
In the FFIFF mode interaction domains Puck depicts the prevailing 3D-state of stress. Additionally, with
applies a weakening factor (depending on  1) reasoning respect to the symmetries of this transversally-isotropic
that single lament failures have a weakening eect on material (modelled an ideal crystal [23,26,34]), the ve basic
the resistance against IFF. Cuntze automatically strengths and ve elasticity properties are given (Lekhnit-
respects this fact by the rounding-o procedure. For skii). A UD-lamina in reality is a low-scale structure with
more information on the dierences and coincidences of the constituents bre, matrix and interphase (at the inter-
Pucks and Cuntzes failure theory, see Appendix A. face). After homogenisation it may be called material.
The theoretical background of the following con-
tribution can also be found in the DURACOSYS 99 3.2. Invariants
paper Progressive failure of 3D-stressed laminates:
multiple non-linearity treated by the failure mode con- Strength criteria or failure conditions may be for-
cept (FMC) [29]. mulated by invariants based on the UD-stresses, see
The authors hope to add, with this lamina stress- [13,30]). Invariants have the advantage that the
based engineering approach, a physically-based 3D formulations do not depend on coordinate-system
phenomenological model. transformations.

Table 1
Main features of the FMC

Each mode represents one theoretically independent failure mechanism and one piece of the complete failure surface (surface of the failure body or
limit surface)
Each failure mechanism is represented by one failure condition. One failure mechanism is governed by one basic strength and therefore has a clearly
dened equivalent stress  eq
Curve-tting of the course of test data is only permitted in the pure failure modes regime
Dierent, however, similar behaving materials obey the same function as failure condition but have dierent curve parameters
Rounding-o in mode interaction zones is performed by the spring model presented.

Table 2
Additional FMC aspects/information

An invariant formulation of a failure condition in order to achieve a scalar potential considering the materials symmetries [34] is possible
Each invariant term of the failure function shall be related to a physical mechanism observed in the solid, causing a volume change or a shape
change or friction
Hypotheses applied:
Hashin/Puck with Beltrami (choice of invariants), MohrCoulomb (friction, thinking in Mohrs stresses)
The rounding-o of adjacent mode failure curves (partial surfaces) in their interaction zone is leading again to a global failure curve (surface) or to a
single surface failure description (such as with Tsai/Wu, however without the well-known shortcomings).
* Proof of design and strength analysis:
For each mode one reserve factor f Res or one stress eort (if nonlinear) is to be determined, displaying, where the design key has to be
The probabilistics-based rounding-o approach delivers the resultant reserve factor linked to the margin of safety by MS=f Res  1.
* Nonlinear stress analysis with Degradation:
Equivalent stresses and stress eorts are used in this (nonlinear) progressive damage description.
Failure mode identication is mandatory for a progressive failure analysis in order to know how the lamina has failed. Criteria which just
predict failure do not make a clear degradation of the moduli possible.
348 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

(adhesive failure caused by a weak bre-matrix bond),

as well as by the bres acting as embedded stress raisers.

3.4. Rounding-o in the interaction zones

Of further interest is the rounding-o of the fracture

curve in the mixed failure domain (MiFD) or interac-
tion or transition zone of adjacent failure modes in the
envisaged lamina. In [24] a simple probability-based
formulathe Series Spring Modelas an engineering
approach for the resultant reserve factor (which is
needed anyway for the proof of design)
1=f Res 1=f mode1
Res 1=f mode2
Res 1=f mode3
Res 2

was proposed which approximates the results of a time-

consuming probabilistic calculation on the safe side. In
Fig. 1. UD lamina (t=tension, c=compression). Stresses, strengths the case of residual stresses and non-linearity instead of
and elasticity properties. a stress-based reserve factor fRes the stress eort E has
to be employed.
From the variety of invariants the following forms
were chosen to best describe the multi-axial behaviour 3.5. Classical laminate theory (CLT)
of the material (the numbering of the invariants is dif-
ferent in the various literature, e.g., I3Boehler I4Hashin and (The CLT is addressed here mainly for the reason to
I5Hashin 2 31
2 2
 3 21 223 31 21 ). depict the denitions and symbols in the German
guideline VDI 2014 on Development of FRP compo-
I1 1 ; Boehler
nents sheet 3: Analysis (issued 2003), chosen after many
I2  2  3 ; discussions, and which will be employed here. Another
2 2 reason is given Section 4.2)
I3 31 21 ;
Assuming transversal isotropy and the state of plane
I4 2  3 2 423
 2  stress (in-plane stressing,  3=0, which is the situation
I5 2  3 31  21  423 31 21 : 1 of the case studies investigated) the linear stress-strain
relations for the kth lamina of a multi-layered laminate
The sensitivity of I5 to the sign of the shear stresses is are (using the notation; 1 k, 2 ?, 12 k?, Q ;
suppressed if a main axes transformation around the [S]:=stiness, compliance matrix of the lamina)
 1-axis is performed (see Fig. 1), leading to  23=0.
f"gk "1 ; "2 ; 12 Tk S k f gk and 3a
3.3. Strengths (Cuntzes view)
f gk 1 ; 2 ; 12 Tk Q k f"gk : 3b
The characterisation of the strength of transversally The symmetric elasticity matrix of stiness (stiness
isotropic composites requires, according to the FMC matrix) of the lamina reads:
the measurement of ve independent basic strengths: 2 3
Q11 Q12 Q16
Rkt , Rkc (bre parallel tensile and compressive strength) as
Q k 4 Q21 Q22 Q26 5
well as R?t , R?c (tensile, compressive strength transversal to
Q61 Q62 Q66 k
the bre direction) and R?k (bre parallel shear strength). 2 3
Rkt is determined by the strength of the constituent Ejj jj? Ejj
bre and Rkc by shear instability. The latter includes 6 1  jj? ?jj 1  jj? ?jj 0 7
6 7
dierent micro-failure mechanisms: The matrix may shear 66 ?jj E? E? 7 ;
7 4
4 1  jj? ?jj 1  jj? ?jj 0 5
under loading and does not stabilise the generally some-
what misaligned bres embedded within. Hence it comes 0 0 Gjj? k
to bending and kinking [41] (structural behaviour). Also,
the load grasping bre as stier constituent may shear with
(this is a constituents material behaviour) under kc and
?k . The strength R?t is determined by the relatively low Q 1
k S k ;
strength properties of the matrix (cohesive failure), by 5
?k  E? k?  Ek MaxwellBetti law
the interphase material in the interface bre-matrix
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 349

and ?k as the major Poissons ratio. Thus, for the and
application of CLT the knowledge of only four con-  0
T k T" k f T gk ; 13a
stants is essential: Ek , E? , G?k and ?k (  12 in failure
In the case of mechanical loading the following load-  T
strain equations are obtained in the cross section for the f T g Tk ; T? ; 0 : 13b
load uxes {n} and the moment uxes {m } (moment
per unit legth of the middle surface) In the case of symmetrical lay-ups (test cases of the

failure exercise), for the treatment of material non-
n A B " " linearity and of degradation, the lamina stresses {}k
K 6
m B D have to be computed considering
" k f"  g . . . compatability 14
with [K] being the stiness matrix of the laminate, from
which will be utilized the extensional stiness matrix
(see Theory of Laminated Plates by Ashton/Whitney  0     
 k Q0 k  "0 k  "0T k . . . Hooke 15
A Q0 k tk ; Q0 T Q T T ; 7
f gk T 1 1
k  fgk ; f"gk T" k f"gk 16

and transformation matrices (s=sin , c=cos ) and applying

2 2 3
c s2 2sc T 1 T" T ; 17a
T 4 s2 c2 2sc 5 8a
sc sc c2  s2 T" 1 T T : 17b
The denitions for the lamina (often called ply if it is
2 2 2
3 a prepreg and layer if it is winding) stresses, angles and
c s sc
thicknesses are illustrated in Fig. 2.
T" 4 s2 c2 sc 5 8b
The index k of the single lamina will be dropped in the
2sc 2sc c2  s 2
further text.

Having determined the strain vector {" } and the

curvature vector { } for the middle plane of the lami- 4. Failure modes and failure conditions of a lamina
nate, the so-called natural strains {"}k (strains in the
lamina coordinate system) and stresses {}k in each Failure conditions [4957] should exhibitbesides a
lamina may be calculated according to sound physical basisthe numerical advantages:
f"gk T 1 
k f" g zf g; 9a

f gk Q k f"gk : 9b

The equations above decouple for a symmetric lay-up

f" g A 1 fng 10

If curing stresses have to be considered the equations

f" g A 1 fng fnT g 11

fn T g DTQ0 k tk 0T k 12 Fig. 2. Laminate and kth lamina subjected to a plane state of stress
k1 (mid-plane z=0).
350 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

mathematical homogeneity (see F? in Appendix A4 [41a]. These micro-cracks grow until they touch the next
after the homogenisation) in the stress terms, stress bre layer where they are turned to later form the basis
terms of the lowest degree, simplicity, scalar formula- for the bre-parallel IFF.
tions and therefore invariance, numerical robustness The explosive eect of a so-called wedge shape fail-
and rapid computation. ure (a ?c -caused IFF) of an embedded lamina of the
laminate may directly lead [42] to nal failure (see a
4.1. Failure modes (types) torsion spring) or via local delaminations to buckling of
the adjacent laminae and therefore to nal failure, too.
A designer has to dimension a laminate versus inter- This IFF, may also cause a catastrophic failure like FF.
bre-failure (IFF) and bre-failure (FF). IFF normally
indicates the onset of failure whereas the appearance of 4.2. Strain energy density basis
FF in a single lamina of the laminate usually marks nal
failure. In the case of brittle behaving FRP, the failure is Beltrami, Schleicher et al. assume at initiation of yield
a fracture. The IFF incorporates cohesive fracture of that the strain energy density will consist of two por-
the matrix and adhesive fracture of the brematrix tions. Thus, the strain energy (denoted by W) in a cubic
interface. element of a material reads
Fracture is understood in this article as a separation
of material, which was free of damage such as technical W f gf"gdf"g WVol Wshape : 18
cracks and delaminations but not free of tiny defects/
aws (size of microns) prior to loading. Including Hookes law in the case of a transversally-
Fig. 3 informs about the types of fracture which are isotropic body the expression will take the shape (see
recognised in case of dense (means: not porous) Lekhnitskii [42a], sik=compliance coecients analogue
transversally-isotropic ideal materials. to the 2D formulation of Eq. (3a). See also Ashton/
Whether a failure may be called a shear stress induced Whitney [41b]):
shear failure, SF, or a normal stress induced normal
failure, NF, depends on the size scale applied. SF?k   2 
W s11 12 s22 22 s33 23 s44 23
s55 12 2
13 =2
shows macroscopically shear failure (fracture plane is
parallel to  21). However micro-mechanically, it is a 45 s12 1 2 1 3 s23 2 3
normal failure mode of the matrix, caused by tensile I12 I 2 1  ?? ?k I1 I2 I3
matrix stress and visualised by the so-called hackles 2 
2Ek 4E? Ek 2Gk?
volume volume volume shape
I4 1 ??

Some of the terms above describe the volume change of

the cubic material element and others its change of the
shape. These changes can be witnessed by the fracture
morphology [41a].
In order to formulate a relatively simple failure con-
dition one chooses as basic invariant that term in Eq.
(19) which respects whether the cubic material element
will experience a volume change in the considered mode
or a shape change.

4.3. Failure conditions achieved

In engineering application due to property scatter the

simplest strength criteria which still describe the physi-
cal eects should be applied. This always reduces the
number of curve parameters (inherent in the failure cri-
teria) to be determined and, besides this, the numerical
Fig. 3. FMC view of the fracture types (  failure modes) of brittle
eort. Applying the FMC in total three (statistically-
transversally-isotropic material. (The physical fracture planes are based) calibration points at maximum have to be
pointed out in the gure [2] /fp :=fracture plane angle). The onset of experimentally determined besides the basic strengths
hackles due to NFm relates IFF2 to IFFI (micro-cracks due to NFm). serving as anchor points in each mode failure domain.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 351

Based on the idea above the following failure condi-  2

?k 1
1  21 =R?k 
tions, Ff g 1 have been derived ?jj 
b?k ?k2 1 3
from 2c ; 21 21a
22c  21 =R?k
FF1 : Fk 1;  
R1 tk 1 2c 3c =R1 c?
b?    2 21b
I1 2c 3c =R1 c? 2c  3c =R1 c2
FF2 : Fk 1;
R1 c k   c  2
p b?k 1  b?  1 2?k =R1 ?k  b? 2?k
=R1 ?k 21c
I2 I4
IFF1 : F? 1 20 for the parameter determination. The parameters
2R1 t ? depend on the material behaviour and on the IFF for-
I2 I3  I5 mulation applied. Bounds on the safe side for GFRP,
IFF2 : F?k I33=2 =R1 3?k b?k 1; CFRP and AFRP are assumed to be
R1 3?k
0:05 < b?k < 0:15; 1:0 < b? < 1:6; 0 < b?k < 0:4:
I2 b? I4 b?k I3
IFF3 : F? b?  1 1
R1 c ? R1 c2 ? The extreme value b?k 0 means no bulge eect and
b? 1 means no friction in the ??-plane. Above
bounds for the parameters and later the mapping of the
with three free curve parameters b?k ; b? ; b?k to be failure curves are based on multi-axial test data cited in
determined from multi-axial test data: (R1 marks mean literature [57] or carried out at MAN.
strength value. *Mind: 1 ! vf  1f vf  "1  E1f "1  The authors experience shows: often, b?k 0 will
Ekt with 1f =tensile stress bre and vf=bre volume map the lamina test data well enough. The skill has to
fraction. The very small load-carrying capacity of the be put into m as a rounding coecient on the safe side.
> Data for the computation of b?jj (Fig. 4) are numerous,
matrix is neglected here in relation to the bres. F1 is
< b?k 0:1 is a good approach. As calibration points for
called criterion). Each of them has to be calculated from b? are still missing in the transversal-isotropic domain
a test point (several measurements) or by curve tting of knowledge from brittle isotropic material is applied
the course of test data in the associated pure domain. which will keep the engineer in the compression domain
The (calibration points & in Figs. 4 and 5 deliver, after on the safe side by assuming b? 1.
inserting them into the equations IFF2 and 3 and a In the following text the reasons are depicted for the
further resolution, the equations application of Which invariant? and of Which form of

Fig. 4. Visualization of the reserve factor and computation example. {}(L)=load stress vector, m =rounding-o exponent [Eq. (45)]. (no curing
352 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

invariant? (success check was the mapping of the avail-

able multi-axial data):

Fk : According to the FMC, Fk originally consists

of a quadratic term in stresses. However, being
the only (basic) term, the quadratic term can be
replaced by a numerically simpler linear term
which regards that the bre tensile stress and not
 1 (the UD material model does not hold here)
has to be applied if formulating a failure condi-
tion. Eq. (20a) indicates that for FF not 1t has to
reach the value for the UD-strength Rkt but "k 
Ek : Why? Poissons eect is not negligible,  
because a compressive lamina stress state 2c ; 3c
will cause tensile bre stress. F  1 theoretically
may be reached even without a load stress 1t !
Fk : Again the basic term is I12 . For reasons of
simplicity and due to lacking of test data in the
2c ; 3c domain, a shear addressing invariant I3
(reecting some Wshape) was not considered in
Fk . By this, the I12 could be reduced to the linear
basic term I1.
F?k : Basic term is I3. The choice of the failure
condition is strongly aected by the easy to be
used desire and by an easy determination of fRes ,
which is simplied if Ff g is a so-called homo-
geneous function wherein the stress terms are of
the same power (grade). Therefore I33=2 =R1 3?k was
applied, instead of a quadratic formulation
which was used in the past, thus leading to
homogeneity of F?k . The term I2 I3  I5 is the
result of an intensive analytical trial and error
search of the rst author. It respects the dierent
interaction of the stress combinations 2 ; 21
and 2 ; 31 a typical material asymmetry at rst
described by Puck and proven by test [6] (not
considerable by Tsai/Wu).A numerical problem
existing in F?k has to be mentioned: If
b?k I2 I3  I5 becomes I33=2 , then the 21 2 -curve
in Fig. 4 turns to innity. In order to generally
bypass this diculty one has to put a query in the
program and replace, if applicable, the formula-
tion of the o-turning F?k curve by a limiting
horizontal line dened by the constant max I33=2
(see Appendix A4). This is very simply done for
the 2D test cases.
F? : After another intensive search the really
straight line in the quasi-isotropic 2 ;p 3 -plane
could be mapped by employing I2 + I4 in F?
(Is the section line with a hyperbola. Known
from isotropy).
Fig. 5. Scheme Global Fit and Mode Fit. Example: CFRPIFF F? : In F? , besides the basic term I4, the linear
curve of UD-material. (MfFD=mixed failure domain=fracture due
term I2 was applied which considers friction. If
to two modes. MfFD=multi-fold failure domain of the same mode
Normal Fracture NF? working twice. A-curve: 99% reliability, really necessary, a term I3 may additionally be
95% condence [34]; B-curve: 0%/95%, mean-curve: 50%/50%). In taken aiming
 at a better numerical rounding-o
A-, B-design space: R1 ! design allowable R. in the F? ; F?k -interaction zone.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 353

On the other hand, if I3 is not applied, that means FoS, j, then a linear elastic modelling is permissible and
sticking to the basic FMC b?k 0, p 
? can be homo-
F a stress-based fRes can be predicted.
genized, too, by replacing I4 =R? by I4 =R?c . This will
In case of non-linear behaviour accurate reserve fac-
lead from a parabola (is in the negative domain already tors have to be referred to loads, which is in accordance
almost straight in Fig. 5for the parabolic formulation) to the fact that load FoS are given. Analysis provides via
to a straight line for 2c 3c , and shall be the authors the failure criterion with the modes equivalent stresses
engineering choice in future (see Appendix A4).  eq and stress eorts E outlining the remaining load
With respect to the 3D character of the IFF condi- capacity for the computation of the resultant reserve
tions above they may serve also as criteria for the onset factor. The value of the reserve factor then is the ratio
of delamination F? : wedge failure, F? : transversal ten-
sile failure) generated by the interlaminar stresses failure load at Eff res
fRes 1:
3 ; 32 ; 31 . Hydrostatic compressive and tensile stressing j  DLL
is automatically considered.
One has to keep in mind: one or two modes will be the As failure load is often taken, the maximum load
design driving ones in a local material point of a com- achieved when computation stops due to numerical
posites lamina. The basic strength of the mode-related problems. Nonlinear analysis in general means stress
linear or non-linear stressstrain curve controls the redistribution in the structure. This lowers the stress
(size) volume of the mode failure surface (body) being level of the hot spots in the laminae (dened material)
one part of the global failure surface (body). Curve of the laminate.
 are representing an eect, such as friction
b? in the material. They control the shape of the mode 5.2. Determination of mode reserve factors
failure surface.
If linear analysis is permitted:

5. Reserve factors f mode res

Res , f Res of the lamina Case No residual stresses: f gL f j  DLL
f gfailure fRes  f gL f gL MS  f gL 23
5.1. General

Reserve factors which have to be determined for the with the margin of safety MS fRes  1.
Proof of Design of each lamina in the laminate are
dened load-related. These are: Inserting the above denition into the failure
for the initial failure, indicated by the so-called  
knee in the laminates stressstrain curve and F Ff gfailure F fRes  f gL 1
originated by F? ; F?k in the laminae
initial failure load yields an equation for the stress-based fRes
f initial
Res 22a
jp0:2  DLL fRes  L f 2Res  qL f 3Res  cL . . . 1:
for the nal failure, indicated by Fk ; Fk or F? ;
final failure load Special example: The failure condition only
f final
Res 22b has linear and quadratic stress terms. Then the
jult  DLL
reserve factor can be calculated [23] by resolving
for fRes as of a polynomial a root which delivers
with DLL=Design Limit Load and jp0.2, jult=design (mind: R1 ! R in the case of Proof of Design
factors of safety (FoS).   
The various failure loads to be inserted into Eqs. (22) fRes 1=L e:g: 1= vf  I1f =Rkt . . . linear
are either a result from experiment or from analysis 24a
(applying a failure criterion).  q
In linear analysis the reserve factor fRes is normally fRes L 2L 4qL =2qL . . . quadr:
dened that factor all mechanical load-induced stresses
applied to the laminae have to be multiplied with in
other to generate failure. Geometrically it means that with L 5 linear terms, qL 5 quadr. terms.
the stress vector f gL has to be stretched in its original
direction by this factor in order to cause failure. This Case With residual stresses (linear modelling)
visualisation is valid as far as linear modelling can be f gfailure fRes  f gL f gR : 25
applied: If there are no residual stresses and high design
354 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

In the case of linear terms, after substitution of achieving a description of the global (complete) failure
the failure causing state of stress one yields surface. This procedure is simple, however error-prone
  in some domains, due to its physical shortcomings.
F Ff gfailure F fRes  f gL f gR
In order to consider failure t probability
 or the multi-
1 26 fold failure
m: in the 2 ; 3 -domain (MfFD) the
term 1=f?Res has to be made 2-fold eective. A sim-
with f gR from curing stresses computation etc. ple numerical way to implement this is by including in
This procedure can be applied as long as the Eq. (27) (see Fig. 5),
residual stresses have not caused an essential
 m:  m:
amount of damage which would lead to stress- via 1=f?
Res 1=f?MfFD ;
redistribution and a reduction of the size of the
residual stresses. the multi-fold failure term [43].
f?MfFd 2R?t = 2t 3t 28
5.3. Determination of resultant reserve factor (rounding-
o of failure modes)
Eq. (28) is applied only, if test data mapping makes it
The (resultant) Reserve Factor (superscript res) takes necessary. The experimental behaviour of brittle iso-
account of the interactions of all modes. In case of tropic materials justies the MfFD rounding in the
linearity it may be estimated (Fig. 4 just includes quasi-isotropic plane of the UD-lamina.
the relevant interacting modes) by the rounding-o In the following set of formula the so-called equivalent
equation or spring model stress of each mode is applied. This stress includes all
 m:   load stresses and residual stresses which are acting
1=f Res f f Res together in a mode equation.
 m:  m:  m: 27
1=f ?
Res 1=f ?k
Res 1=f ?
Res 5.4. Application to the UD-lamina (3D-conditions)
 m:  m:
1=f k
Res 1=f k
Res The Mode Reserve Factors explicitly read
: generally f Res R mode =eq
; 29
with m as the rounding-o exponent, which ts test
data. As a simplifying assumption m is taken the same  
: f k
for each interaction zone. The value of m has to be set ^ Rkt = "1  Ekt Rkt =eq
; 30a
by tting experience and by respecting the fact that in
the interaction zones micro-mechanical and probabilistic f k c c k
Res Rk =1 Rk =eq ; 30b
eects will commonly occur and cannot be dis-
criminated. From numerical reasons m should be an 2R?t R?t
odd number between 2.5 and 4. f ?
Res p ? 30c
I2 I4 eq
If inserting a unidirectional fracture stress (this is the
strength value) into the equation above, then a failure curve R?c
f ?
or a failure surface described by f res
Res 1 is achieved.

Fig. 5 refers to the 2 ; 3 -plane as one failure plane     2
b?  1 I2 b?  1 I22 4b? I4 4b?k I3
of the various ones. In the upper part it visualizes the 
evaluation of test data and in the bottom part the b? I4 b?k I3
rounding-o (by the spring model) in the multi-fold 30d
(MfFD) and mixed failure domains (MiFD) as well as  1=3
the shrunk design space (mean strength R1 of mapping is f ?k
Res R?k = I33=2 b?k I2 I3  I5 30e
replaced by a strength design allowable R) to be used by
the designer in the dimensioning and in the proof of
design. The rounding shown in the Figs. 4 and 5 seems Remark. If f Res becomes negative, caused by the
to exclude the FF modes. These modes, however, have numerically advantageous automatic insertion of f g
no relevant interaction with the failure curves 21 2 1 ; 2 ; 3 ; 23 ; 31 ; 21 T as FEM output into Eq. (30), a
and 2 3 . value of +100 shall replace the negative value. A nega-
Additional to the FMC-based Mode Fit the Global tive value e.g. results if a positive  1 (better "t1 Ek ) is
Fit (e.g. Tsai/Wus single failure surface criterion inserted into Eq. (30b).
describes a global failure surface or body) is pointed For an eective design the stress engineer is to be
out. The Global Fit interacts between the UD-stresses provided with a table which indicates the design driving
and includes independent failure modes in one equation, mode reserve factors (an example: see Appendix A3).
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 355

6. Equivalent stress, mode eort and eective secant moduli for decreasing stress (Softening) 7 < 0
<mode mode
correq eq =TrF 35b
In the case of small FoS (e.g. in spacecraft) just non-
linear analyses will enable the stress engineer to predict
the stress eort and then the load-based fRes . The actual being a modulus decrease, with the trigger-factor
stress eort of a mode, Eff mode , is the actual portion of
TrF Eff res =max Eff mode : 36
the maximum 100% achieved at mode failure. The pro-
cedure of determining the resultant stress eort
Eff res in each lamina of the laminate is similar to that In these equations the stress eort of the maximum
of f Res : The stress eort (Puck calls it stress exposure stressed mode governs the triggering and TrF is dedi-
factor fE ) can be related to the reserve factor in case of cated to all IFF modes aected. As Eq. (36) leads to a
linear behaviour and zero residual stresses, that means sharp decay,p: a damped triggering according to
on stress level, by newTrF m TrF is proposed for the future.
res This approach has to be veriedbefore general
Eff res 1=f Res : 31
acceptancefor all possible stress combinations, of
course (see also Appendix A1.2).
Also similar to the f Res procedure at rst the
equivalent stress vector
n o  T 7. Description of non-linearity
modes k k ? ? ?k
equiv: eq ; eq ; eq ; eq ; eq 32
Non-linear behaviour [5860] of well-designed com-
will be computed. It includes the equivalent stress of posites is most often physically (laminae behaviour) but
each mode of the lamina and within the nonlinearly rarely geometrically (laminate behaviour) caused.
load dependent load stresses f gL and the equally non- A full 3D-input in stress analysis demands for ve
linearity-dependent residual stresses f gR from curing elastic properties in the case of Fibre Reinforced Plastics
etc. (FRP) and in strength analysis for ve strengths. In the
Consequently the resultant stress eort is represented 2D-case the required input consists in four elastic prop-
by erties and ve strength properties.
: X5 Further, for the non-linear stress analysis the relevant
Eff res Eff modes non-linear stressstrain curves are to be provided, which
1 should discriminate the so-called hardening and the
 m:  m: softening (Fig. 6). Material hardening is the domain
k 1 t k 1 c
eq =Rk eq =Rk until the stress reaches its strength value Rm which
 m:  m:  m: addresses here an initial failure level of IFF type. From
? 1 t ?k 1
eq =R? eq?
=R?c eq =R?k 33 that level on, that means for the progressive failure or
damage regime, the term softening is used. Of course,
with Eff modes corresponding to some extent to the some damaging already begins with material hardening.
f Edomains of Puck (see Appendix A).
In case of fracture stresses holds, analogous to f res
Res 1;

Eff res 1 100%: 34

Usually in the laminae of a laminate, multi-axial
states of stress are acting which have an impact on more
than one of the failure modes. Because in the interaction
zones adjacent failure modes are commonly aected, a
corresponding degradation (displayed by a stiness
reduction) has to be considered by a drop in the secant
moduli applied in the non-linear analysis. A triggering
of the adjacent equivalent stresses takes into account
this eect for each of the associated moduli. As triggering
approach is recommended (see also Appendix A1.2):

for increasing stress (Hardening) 7 > 0

>mode mode
correq eq TrF 35a
Fig. 6. Mapping of measured stressstrain of an isolated UD-speci-
being a modulus decrease mens. [Eqs. (37) and (40)]. [Example t21(g21)].
356 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

7.1. Mapping of hardening

The degree of non-linearity essentially depends on the

nonlinearly behaving matrix material which aects E?c
and Gjj? . For the secant moduli to be applied in the
non-linear stress analysis the following values are
determined by the Ramberg/Osgood equation which
maps the course of non-linear stressstrain data very
well (with E(o) the initial tangent modulus)
" =Eo 0:002 =Rp0:2 37

with the Ramberg/Osgood exponent (see [5])

n n "pl Rm =n Rm =Rp0:2 38
estimated from the strength point Rm ; "pl Rm . Data for Fig. 7. The dierences in the stressstrain behaviour of isolated and
the secant moduli of E? , Gjj? are provided from above embedded UD-laminae [for the (b) and (c) curve Eq. (40) is applied.
Ramberg/Osgood mapping of test data course (denota- The softening parameters for (b) and (c) are dierent].
tions see Fig. 6) by
  n1  with two curve parameters as, bs to be estimated by the
Esec Eo = 1 0:002  Eo =Rp0:2  =Rp0:2 39 data of two calibration points, e.g.
Rm ; "Rm and Rm  0:5; "Rm  0:5: 41
7.2. Mapping of softening
The above softening function [Eq. (40)] practically
Above the Initial Failure level an appropriate pro- models the stressstrain curve of a lamina which is
gressive failure analysis method has to be employed (or embedded in a laminate, and thus, it includes the eect
a Successive Degradation Model for the description of of the altering microcrack density up to the critical
post initial failure) by using a failure mode condition damage state (CDS). Curve (c) is therefore an effective
that indicates failure type and damage danger (level of curve.
stress eort). Final Failure occurs after the laminate
(and thereby the structure) has experienced a stiness 7.3. Constraint eect on an embedded lamina
reduction and has degraded to a level where it is no
longer capable of carrying additional load. If applying test data from tensile coupons to an
Fig. 6 depicts hardening with softening. In detail: for embedded lamina in a laminate, one has to consider that
an isolated e.g. tensile coupon specimen (a) in the usual tensile coupon tests deliver test results of weakest link
load controlled test, (b) in a strain controlled test. A type (series model). An embedded [40] or even an only
measurement of curve (b) would be possible at the one-sided constraint lamina, however, belongs to the
institute BAM in Berlin, which possesses a MTS test rig class of redundant type behaviour, to the parallel spring
of a very high frame stiness (load goes via the sti model type. Due to being strain-controlled the material
frame and not via the relatively weak specimen), how- aws in a thin lamina cannot grow freely up to micro-
ever, tests have not yet performed. The curve (b) is crack size in thickness direction, because the neigh-
assumed here due to the lack of experimental data from bouring laminae will act as micro-crack-stoppers (prob-
there. Modelling of Post Initial Failure behaviour of a lem of energy release in fracture mechanics).
laminate requires that assumptions have to be made Cuntze sees the peak value of so-called eective
regarding the decaying elastic properties of the actually stressstrain curve (in situ, embedded lamina) slightly
degrading embedded lamina [curve (c) in Fig. 7]. E?c and higher than the strength point R1 of the isolated speci-
Gjj? are decreasing gradually rather than being suddenly men due to the change from the weakest link beha-
annihilated. A rapid collapse (often named ply discount viour to the real redundant behaviour (Fig. 7) of a
method) of E?t is unrealistic and probably further laminate. For the sake of simplicity this peak value is
leads to convergence problems. A simple function was lowered down to R1 in the following analytical descrip-
used to map this softening, in order to later derive the tion of softening. For the execution of non-linear ana-
secant moduli. It generally reads (the sux s denotes lysis the application of an eective stressstrain curve is
softening) necessary which estimates the behaviour of the lamina
in the laminate regarding the stack, its position, and the
s Rm =1 expas "=bs 40 thickness.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 357
In order to provide the non-linear analysis with the formula. Due to the fact that 1 1 dierent inter-
input needed, normalized stressstrain curves have been action eects can be accounted for. A recommendation
constructed (Fig. 8) with a hardening part measured of the author for an improved treatment of the micro-
and a softening part assumed (as long as no test data are mechanically linked modes F? andF?jj is derivable from
available).  m: 1  m: 3  m: 4
In the non-linear analysis normally mean values have 1 1=f jj
Res . . . 1=f ?
Res 1=f ?jj
Res ::: 42
to be regarded in order to perform a stress analysis that
corresponds to an average structural behaviour. There- Utilizing dierent exponents the solution has to be
fore, when executing a non-linear stress analysis of the achieved iteratively :
structure by a mean stressstrain curve, the secant
res1 res1 res2
moduli to be utilized are mean values, too. However fgL ! f Res ; f Res  fgL ! f Res ;
later, in the strength analysis (Proof of Design) of the
res2 res1 res3 resj
hot spots so-called A or B design allowables [44] as f Res  f Res fgL ! f Res until f Res  1:
minimum strength values R (no bar upon R) have to be
res res1 res2
regarded. Hence will be f Res f Res  f Res   .
For simply deriving clear data for the secant moduli The procedure for the stress eort E(res) is
two regimes have to be distinguished: one below and analogous.
one above "R1 m .
7.5. Variation of Poissons ratio
7.4. Choice of dierent m values
The alteration of the major Poissons ratio ?k (nota-
In the rounding-o or interaction equation just one tion VDI 2014 [23a]) is linked to the associated failure
constant value for m is inserted. This might not work if mode. For example, in the case of shear failure under
the interaction eects covered by rened conditions (e.g. compressive lateral stresses the value for ?k will be
[23]) are replaced by more practicable simpler formula- higher than for tensile lateral stresses. Respecting the
tions, (e.g. setting b?jj =0, Appendix A4). In that case low eect Poissons ratios haveif using FRP with sti
the rounding-o equation may be split into several bresthe following estimation will be a good
mode interaction formulae replacing the single equa- approach before mode failure occurs:
tion, because interaction addresses two or at maximum
four of the ve modes, only. The advantage of this F? : ?jj ?jj0  E?sec =E?0 :
computing intensive procedure would be the possibility
of accounting for dierent values with respect to dier- Also in the case of F?k the value for ?k is reduced.
ent interaction eects in the various mode interaction
zones. 7.6. Remarks on design and modelling
If the failure curve is reached, then f Res 1, and for
this level one can stay with the advantage of one single In composite structures composed of sti bres
and hopefully well-designed by netting theory the
bre net controls the strain behaviour.
The FMC considers the inter-laminar stresses
and classies the failure modes. Therefore, asso-
ciated degradation models are inherent and make
a gradual degradation of the aected property
In order to design a laminate properly, not only
veried failure conditions have to be available,
but also proper stresses have to be analytically
provided[45]. Therefore, analogous to isotropic
materials, the non-linear stressstrain curves
have to be taken into account below reaching the
initial failure level.
Above the initial failure level an appropriate
progressive failure analysis method has to be
employed by taking a Successive Degradation
Model and by using a failure mode condition that
indicates failure type and quanties damage
Fig. 8. Normalized stressstrain curves. danger or fracture risk.
358 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

Final failure occurs after the structure has 7 > 0 (increasing stress, hardening)
degraded to a level where it is no longer capable
of carrying additional load. This is most often E t?sec E t?o
caused by FF, however in specic cases by IFF,  
too. An inclined wedge-shaped inter-bre crack E c?sec E c?o = 1 0:002  E c?o =Rp0:2

caused by F? can lead to nal failure [42].  n? 1 c

Multidirectional laminates are usually still cap- ? ?c
 eq =Rp0:2
able of carrying load beyond initial failure which  
usually is determined by IFF. ?jj
Gjj?sec Gjj?o = 1 0:002 Gjj?o =Rp0:2
 n?jj 1
?jj ?jj
 eq =Rp0:2 43
8. Calculation procedure

Fig. 9 presents a suitable ow chart of the non-linear 7 < 0 (decreasing stress, softening)
calculation. The solution procedure of the non-linear  
analysis aims to establish static equilibrium on each E t?sec eq
? ?
=" eq
load step after material properties have been changed. " ! #
  R?t  eq
? ?t s
For each iteration the procedure is repeated until con- eq =bs = n ?
 ?t : 44
vergence (equilibrium) is reached or total failure. A eq bs
correction of the bre angle in accordance with the
change of the specimens geometry has been considered. For the further modes the same formula is valid,
By employing the equivalent stress reached in each fail- however, the mode parameters are dierent. Eq. (44)
ure mode the associated secant modulus of each mode may be transferred to Pucks degradation function Z
was determined for the hardening and the softening (see also Appendix A1.2). After having reached
regime. E(res)=1 this value is kept in the further degradation
Considering a consistent stress concept for all eq procedure which causes a stress redistribution towards
an explicit dependency for Esec eq has to be pro- the bres as far as the bre net allows it. Thereby, also
vided. For reasons of achieving such an explicit for- the residual stresses are reduced similar to the situation
mulation two separate formulae are discriminated with metallic materials where increasing non-linearity
which are linked in the strength point. This auto- reduces stiness, and, the residual stresses.
matically respects that the chosen non-linear calculation If the laminates stiness matrix is recomputed after
procedure demands for the dependencies of the secant each step of damage increase the laminates damage
moduli on the corresponding equivalent stress. These evolution may be continuously monitored. The approach
dependencies are: may be called a self-correcting secant modulus procedure.

9. Application to test cases

9.1. Denition of test cases

In Tables 35 the mechanical and thermal properties

for laminae and its constituents bres and matrices are
presented. Table 6 provides a survey of the required
plots and the associated loading conditions. The plots
are nonlinearly to be computed.

9.2. Assumptions and remarks for the plots

Post-initial failure is considered by gradually
degraded properties of embedded laminae (no
Sudden Death of the failed lamina). The course
of the softening (sux s) is assumed
First FF is nal failure. The two FF Fjj (tensile
bre failure) and Fjj (shear instability, local
buckling), and sometimes the IFF Fjj , are dened
Fig. 9. Non-linear calculation scheme (chosen). to cause nal failure
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 359

Failure mode identication according to Cunt- as eective temperature dierence (Table 3) is

zes denition is inherent to the Failure Mode applied (if to be regarded) in order to consider
Concept the eect of curing stresses (are thermal residual
Parameters m; b? ; b?jj ; and b?jj are roughly stresses of the 1st kind).
assumed for the given UD-test cases Moisture may be assumed here to have a bal-
Comment: as temperature drop the dierence ancing eect of 30  C. Chemical shrinking [6] and
stress free temperature minus room temperature thermal curing stresses do not aect the shear

Table 3
Mechanical and thermal properties of the four UD-laminae of the failure exercise [Sod98]d

Fibre type AS4 T300 E-glass 21xK43Gevetex Silenka E-Glass1200tex

Matrix 3501-6 ep. BSL914C ep. LY556/HT907/DY063 epoxy MY750/HY917/DY063 epoxy

Specication Prepeg type Filament wind. Filament wind. Filament wind.

Manufacturer Hercules DFVLR DLR DRA
Fibre volume fraction, Vf 0.60 0.60 0.62 0.60
Longitudinal modulus (GPa) E|| 126a 138 53.48 45.6
Transverse modulus (GPa) E? 11 11 17.7 16.2
In-plane shear modulus (GPa) G||? 6.6a 5.5a 5.83a 5.83a
Major Poissons ratio  0.28 0.28 0.278 0.278
Through thickness Poissons ratio ?? 0.4 0.4 0.4 0.4
Longitudinal tensile strength (MPa) Rt|| 1950b 1500 1140 1280
Longitudinal compressive strength (MPa) Rc|| 1480 900 570 800
Transverse tensile strength (MPa) Rt? 48 27 35 40
Transverse compressive strength (MPa) Rc? 200b 200 114 145b
In-plane shear strength (MPa) R? || 79b 80b 72b 73b
Longitudinal tensile failure strain (%) et|| 1.38 1.087 2.132 2.807
Longitudinal compressive failure strain (%) ec|| 1.175 0.652 1.065 1.754
Transverse tensile failure strain (%) et? 0.436 0.245 0.197 0.246
Transverse compressive failure strain (%) ec? 2.0 1.818 0.644 1.2
In-plane shear failure strain (%)  ? || 2 4 3.8 4
Strain energy release rate (J m2) GIC 220c 220 165 165
Longitudinal thermal coecient (106/ C) || 1 1 8.6 8.6
Transverse thermal coecient (106/ C) ? 26 26 26.4 26.4
Curing: Stress free temperature ( C) 177 120 120 120
2 h at 120  C 2 h 90  C, 1.5 h
2 h at 150  C 130  C, 2 h 150  C
(Eective temperature dierence ( C)e 125e 68 68 68 )
Initial modulus.
Non-linear behaviour, stress/strain curves and data points are provided.
Double cantilever specimen.
Assumption: linearized, reference temperature=RT=22  C.
177+RT+30 (with moisture ef-fect,which is never to be considered in the exercise)=125  C. Temperature drop :=Stress free temperature
minus RT (! just curing stresses: are considered, where to be considered).

Table 4
Mechanical and thermal properties of the four bres utilized [Sod98]

Fibre type AS4 T300 E-glass 21K43, Gevetex Silenka E-Gl.1200tex

Longitudinal modulus (Gpa) Ef|| 225 230 80 74

Transverse modulus (GPa) Ef? 15 15 80 74
In-plane shear modulus (GPa) Gf||? 15 15 33.33 30.8
Major Poissons ratio f? || 0.2 0.2 0.2 0.2
Transverse shear modulus Gf||? 7 7 33.33 30.8
Longitudinal tensile strength (MPa) Rtf|| 3350 2500 2150 2150
Longitudinal compressive strength (MPa) Rcf|| 2500 2000 1450 1450
Longitudinal tensile failure strain (%) etf|| 1.488 1.086 2.687 2.905
Longitudinal compressive failure strain (%) ecf|| 1.111 0.869 1.813 1.959
Longitudinal thermal coecient (106/ C) Mf|| 0.5 0.7 4.9 4.9
Transverse thermal coecient (106/ C) Mf? 15 12 4.9 4.9
360 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

stresses. Micro-mechanical curing stresses (resi- In respect of the few multiaxial lamina test data
dual stresses of the 2nd kind at lament/matrix one single value m=3.1=const will be taken for
level) could not be assessed and are not con- the various test cases.
sidered. They are usually assumed to be respected
in the values for the UD-strengths, For the computation of the test cases the following
The given stressstrain curves of the UD-lamina failure conditions will be employed ( 3 is included only
are interpreted mechanical load-based macro- in the equations where they are eective):
mechanical stresses. It is assumed that the stress
"1  Ejjt 1
strain curves are mean curves (R1 -values are FF1; 2 : 1; 1
given), the curve type one needs for test data Eff jj  Rjjt Eff jj  Rjjc
mapping (see Fig. 5) 2  3 b?jj 22 21 2
IFF1; 2 : ? t 1; 21 3 1
An edge eect (3D state of stress) is not con- Eff  R? Eff ?jj  R?jj
sidered, because the laminates are assumed to be   
b  1 2 3 b? 2  3 2 b?jj  21 2
part of a closed composite structure IFF3 : ? ?  2 1
A progressive behaviour of Ejjt (valid for C-bres, Eff  R?c Eff ?  R c ?
only) was not regarded (see Fig. 15) 45
The loading is monotonic and proportional. No
loading path eects are considered (should be Herein  3=pex is to be inserted in the case of tube
considered some time) specimens loaded by external pressure pex,. For at

Table 5
Mechanical and thermal properties of the four matrices utilized

Matrix type 3501-6 ep. BSL914C ep. LY556/HT907/DY063 epoxy MY750/HY917/DY063 epoxy

Manufacturer Hercules DFVLR Ciba Geigy Ciba Geigy

Longitudinal modulus (GPa) Em 4.2 4.0 3.35 3.35
In-plane shear modulus (GPa) Gm 1.567 1.481 1.24 1.24
Major Poissons ratio m 0.34 0.35 0.35 0.35
Longitudinal tensile strength (MPa) Rtm 69 75 80 80
Longitudinal compressive strength (MPa) Rcm 250 150 120 120
In-plane shear strength (MPa) Rm 50 70
Longitudinal tensile failure strain (%) etm 1.7 4 5 5
Longitudinal thermal coecient (106/ C) Tm 45 55 58 58

Table 6
Summary of laminate types, material types and plots required from contributors (^ x nx =t ; ^ y ny =t; t=laminate thickness)

Laminate type Material type Plots required and description ooading conditions

0 unidirectional lamina (isolated) E-glass/LY556/HT907/DY063T300/BSL914C 1.  2 vs  21 failure stress envelope

E-glass/MY750/HY917/DY063 2.  1 vs  21 failure stress envelope
3.  2 vs  1 failure stress envelope
[90/+30/-30]s laminate E-glass/LY556/HT907/DY063 4. ^ y vs ^ x failure stress envelope
t=2.0 mm, t90=.0.172, 5. ^ x vs ^ xy failure stress envelope
[90/45/45/0]s laminate AS4/3501-6 (quasi-isotropic, widely used) 6. ^ y vs ^ x failure stress envelope
t=1.1 mm, tk=t/8 7. Stress/strain curves under uniaxial tensile
loading for ^ y =^ x =0/1
8. Stress/strain c. for ^ y =^ x =2/1
[+55/55]s angle-ply laminate E-glass/MY750/HY917/DY063 9. ^ y vs ^ x failure stress envelope
t=1.0 mm, tk=t/4 (piping, pressure vessels) 10. Stress/strain curves under uniaxial tensile
loading for ^ y =^ x =1/0
11. Stress/strain c. for ^ y =^ x =2/1
[0/90]s cross-ply laminate E-glass/MY750/HY917/DY063 12. Stress/strain curve under uniaxial tensile
loading for ^ y =^ x =0/1
t=1.04 mm, tk=t/4
[+45/45]s angle ply laminate E-glass/MY750/HY917/DY063 13. Stress/strain c. for ^ y =^ x =1/1
t=1.0 mm, tk=t/4 14. Stress/strain c. for ^ y =^ x =1/1

Note: There is a discrepancy between Sodens Table 11 and his text [3]. This causes no dierence for plot 7 but for plot 10. The latter was corrected
to 1/0 according to Sodens text.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 361

specimens holds  3=0. The consideration of  3=pex In order to dierentiate between the various stress
shifts the biaxial strength capacity to higher values. In terms the following denition is given:
the modes IFF1 and 2 the pressure  3=pex has no
eect. For the computation of the stress eort the Stresses: load stresses, hygro-thermal stresses, residual
particular 2D-state of stress ( 1, 2, 21) has to be stresses
inserted into Eq. (45). This will either not lead to failure, Curing stresses from chemical shrinkage, mismatch of
if E(mode) < 1, or to failure if E(mode) is exceeding the coecients of thermal elongation
value 1. Residual stresses: actively or passively built-in stresses
The modes IFF1 and IFF2 may be called harmless from curing, pre-stressing.
failures whereas IFF3 may cause a catastrophic failure
which is respected in the non-linear analysis.
The equivalent stress, building up the denominators, 9.3. Stressstrain curves of the UD-lamina
was dened by
In the following Figs. 1017 the course of the test data
Eff  R eq f g; 46 (solid lines) is displayed as well as the softening curve
which is assumed for the embedded UD-lamina (dotted
including the residual stress by a superposition to the curve). One remark has to be added here: The dotted
load stress according to part of the F? -curve (Figs. 10 and 14) is only active if
catastrophic failure of the delamination initiating
f g f gL f gR : 47 wedge (its oblique micro-cracks are still closed yet deli-
ver some compliance) is prevented by the laminate.

The residual stresses in the lamina of the laminate are

decaying with decreasing stiness caused by the degra-
dation which accompanies increasing non-linearity. In
other words: in parallel to the decay of the stiness the
non-linear analysis sets matrix-dominated stresses free:
These include thermal residual stresses (curing stresses),
thermal stresses, and mechanical stresses across the
bres. Further, the reduction of matrix-dominated
stresses directly is followed by a reduction of the matrix
stresses balancing bre stresses.
This fully holds for curing stresses of the 1st kind
(upper or material level). A reduction of the not-
respected curing stresses of the 2nd level (bre-matrix
level) also takes place, however, less pronounced.
Curing stresses are respected for laminates, only (on Fig. 11. Transv. compr. stressstrain curve 2c "2 ; UD-lamina (soft-
ening parameters assumed). CFRP: E-glass/MY750/HY917/DY063
1st kind level). [3]. R1 c? =145 MPa, E1 c?0 =16.2 GPa; n?c =6.6; a?c s =3.45%,
s =0.47%.

Fig. 10. In-plane shear stressstrain curve  21( 21); UD-lamina (soft-  
ening parameters assumed, no curing stresses). GFRP: E-glass/ Fig. 12. Transv. tensile stress/strain curve 2t "t2 UD-lamina (soft-
MY750/HY917/DY063 [3]. R1 ?k =73 MPa, G1 k?0 =5.83 GPa; n?k =6.6; ening assumend). CFRP: AS4/3501-6 epoxy [3]. R1 t? =48 MPa,
a?k ?k
s =7.0%, bs =0.53% (assumed). E1 t?0 =11 GPa; a?t ?t
s =1.2%, bs =0.15%.
362 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

9.4. Biaxial failure envelopes for the UD-lamina

In the following UD failure envelopes the residual

stresses are not regarded. Thus, only the so-called load
stresses from the mechanical load test are considered.
For the non-linear analysis the Ramberg/Osgood expo-
nent and the assumed softening parameters of Eq. (40)
are added to each capture.
The course of the presented test curves has been ver-
ied by tests at MAN and tests cited in literature [6,46].
Figs. 5 and 1822 depict several cross-sections of the
ve-dimensional IFF-body:

Fig. 13. In-plane shear stressstrain curve  21( 21); UD-lamina (soft-
Fig. 5: in the graphs ( 2, 3) and ( 23, 2), the
ening parameters assumed). CFRP: AS4/3501-6 epoxy [3]. R1 ?k =79 latter was not shown here, fracture may be excel-
MPa, G1 k?0 =6.6 GPa; n?k =5; a?k ?k
s =4.0%, bs =0.46%. lently described by the homogenized stresses.
Fig. 18: the graph ( 21, 2) represents the IFF-
responsible stresses in the plane of the lamina; the

Fig. 14. Transv. compr. stressstrain curve UD-lamina (softening

parameters assumed). CFRP: AS4/3501-6 epoxy [3]. R1 c? =200 MPa, Fig. 16. In-plane shear stressstrain curve  21( 21); UD-lamina
E1 c?0 =11 GPa; n?c =5; a?c ?c
d =2.7%, bd =0.12%. (degradation assumed). CFRP: R1 ?k =72 MPa, G1 k?0 =126 GPA;
n?k =5; a?k ?k
s =7.0%, bs =0.54%.

Fig. 17. In-plane shear stressstrain curve  21( 21); UD-lamina

Fig. 15. Longit. tensile stress/strain curve 1t "1 UD-lamina. CFRP: (degradation assumed). CFRP: T300/BSL914C epoxy [3]. R1 ?k =80
AS4/3501-6 epoxy [3]. R1 t?k =1950 MPa, E1 tk0 =126 GPa. MPa, G1 k?0 =5.5 GPa; n?k =5; a?k ?k
s =7.0%, bs =0.53%.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 363

graph ( 31, 2) outlines that  31 does not have the

same action plane as 2t (at rst investigated by
Puck, not derivable in Tsai/Wus approach).
Fig. 19: the graph ( 2, 1) shows the limited
applicability of the homogenized lamina stresses,
because  1 or I1 is not the fracture stress. This is
the bre stress  1f. In order to maintain the
composite level in the graph the bre stress is
multiplied by the bre volume fraction
(approach: 1f  vf "1  Ejjt ).
Figs. 20 and 21: in the graph (2=3,1) the
pecularities of a 2D lateral stressing are depicted.
In the domain 2c 3c > 10R?c failure is
caused, not by IFF, yet due to Poissons eect by
Fjj . The zoom, visualizes the rounding-o in one
interaction domain (F? =Fjj ).
Fig. 22: this graph eventually highlights the
( 21, 1)-interaction.

Fig. 20.
p Biaxial failure stress envelope 2 3 ; vf  1f ; ?? 
R?t =m 2 [23]. in MPa. UD-lamina E-glass/MY750 epoxy [3].
b? =1.56, b?k =0.12, b?k =0.4, m=3.1 [Eq. (45)].

Fig. 18. Biaxial failure stress envelope ( 21, 2); and ( 31, 2) in MPa.
UD-lamina (no curing stresses). GFRP: E-glass/LY556 epoxy [3].
Assumed: b? =1.5, b?k =0.13, b?k =0.4, m=3.1 (further data, see
[5,6]). [Eq. 45]. No. 1 of Table 11 [3].

Fig. 21. Zoom of Fig. 20.

Fig. 19. Biaxial fail. stress envelope ( 2, vf 1f) in MPa. UD-lamina. E- Fig. 22. Biaxial fail stress envelope  21(Vf  1f) in MPa. UD-lamina
glass/MY750 epoxy [3], [Eq. (45)]. b? =1.56, b?k =0.12, b?k =0.4, T300/BSL914C epoxy [3]. b? =1.53, b?k =0.15, m=3.1. See also [46].
m=3.1 [further data: see e.g. [46],No. 3 of Table 11 [3]. [Eq. (45)] No. 2 of Table 11 [3].
364 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

9.5. Initial and nal biaxial failure envelopes pressure pex the multi-axial strength is increased
( 3=pex is acting in a favourable manner). The sliding
For the determination of the failure envelopes (see friction due to pex is increased similarly until its max-
Figs. 2326) the code Mathcad, non-linear CLT, and an imum will be reached.
assumed softening behaviour were applied. The symbols Mind: a correct analysis of boundary conditions and
used to indicate the mode of failure are the symbols stress state of the test specimen is mandatory before
which characterize the failure function, e.g. ?k for Fjj evaluating and applying the data.
and so on. The angle marks the associated lamina.
Fig. 24 depicts the symmetrical failure envelopes
Fig. 23 incorporates the initial and the nal of this CFRP laminae. The sharp corners still
failure envelope of this GFRP-laminate. have to be rounded-o in a rened procedure
taking into account the joint failure probability
In the positive quadrant there are no corners. Gen- of the laminate [38]. In the negative quadrant IFF
erally, corners become smoothed due to the eect of covers FF.
high interaction of the failure modes. In the domain The last two failure envelopes (Figs. 25 and 26)
AB both Fjj in the two adjacent laminae are acting are concerning
 the [90/+30/30]s-laminate
together. jected to a ^ x ; ^ y state of stress and a ^ xy ; ^ x
In the negative quadrant wedge failure may occur in state of stress. Again here, sharp corners should
the compressed laminate specimen. The event of a be rounded there where the joint failure prob-
wedge failure is equal to the onset of delamination ability of the failure modes comes to act.
damage. In case of a plane laminate specimen, despite
the anti-buckling device applied when testing in the
compression regime, the wedge will slide and then cause
a compressive reaction  c3 normal to the laminas plane
onto the adjacent laminae (see Pucks drive shaft [42]).
This will induce delamination or might increase an
initial delamination size. However, in case of a pressure
loaded tension/compression-torsion tube specimen (applied
at MAN; see also [6]) and in case of high pressure vessels
(1000 bar, ARIANE 5 launcher) loaded by external

Fig. 24. Initial and nal failure envel. ^ y ^ x in MPa. [90/+45/45/

0]s-laminate, AS4/3501-6 [3]. [Eq. (45)] No. 6 of Table 11 [3].

Fig. 23. Initial and nal failure envel. ^ y ^ x [+55/55]s-laminate, E-

glass/MY750 epoxy [3]. b? =1.5, b?k =0.13, b?k =0.4, m=3.1 [Eq. (45)] Fig. 25. Initial and nal biaxial failure envel. ^ xy ^ x [90/+30/30]s-
^ y =average hoop stress of the laminate, x=0 direction. Limit of laminate. E-glass/LY556 epoxy. No. 5 [3]. m=3.1. [Eq. (45)] ^ x is
usage (lou) at =4%. Curing stresses not included. No. 9 of Table 11. 
parallel to 0 -direction. No. 5 of Table 11 or of Table 6, at hand.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 365

9.6. Stressstrain curves of the laminates design limit or limit of usage (lou) was assumed here to
be max=4% shear strain which corresponds more or
The following stressstrain curves (Figs. 2733) con- less to the shear fracture strain of the isolated lamina.
sider Eq. (45) and the data from Tables 3 and 6 [3]. The As the authors were asked to provide the text with
loading is monotonic, a temperature drop from curing more test data, Fig. 34 was added.
(causes an o-set) is regarded.
Figs. 27 and 28 outline the deformation behaviour of
a pressure vessel, which is usually designed for one spe- 10. Some conclusions, outlooks
cial load case inner pressure that means for
^ y =^ x 2:1. Load combinations outside of this ratio A general concept was highlighted for the
such as 1:0 (Fig. 27)will lead to too high shear strains establishment of Failure Conditions (F=1) for
and thereby to a limit of usage (lou). This shear strain Initial Failure (corresponding to IFF) of dense,
brittle laminae and Final Failure of the laminate.
The complete failure surface consists of piecewise
smooth regimes (partial failure surfaces). Each
regime represents one failure mode and is gov-
erned by one basic strength

Fig. 26. Initial and fmal biaxial failure envel. ^ y ^ x [90/+30/30]s-

laminate. E-glass/LY556 epoxy [3]. [Eq. (45)] No. 4 of Table 11 [3] or Fig. 28. Stressstrain curves ^ y :^ x =2:1 [+55/55]s-laminate. E-glass/
of Table 6, at hand. MY750 No. 11 of Table 11 [3].

Fig. 27. Stressstrain curves ^ y :^ x =1:0. [+55/55]s-laminate, E-

glass/MY750[3]; b? =1.5, b?k =0.13, b?k =0.4, m=3.1. max =4%. Fig. 29. Stressstrain curves ^ y :^ x =2:1. [90/+45/45/0]s-laminate.
[Eq. (45)] No.10 of Table 11, see [3]. AS4/3501-6 epoxy. No. 8 of Table 11 [3].
366 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

Sucient for pre-dimensioning are the basic

strengths R. The remaining unknown curve
parameters b?jj ; b? ; b?jj can be estimated if test
data are not available. The rounding-o expo-
nent m , after some tting experience, can be xed
on the safe side by taking a little lower value.
The interaction (rounding-o) of adjacent failure
modes is automatically considered when calcu-
lating the stress eort Eff res as function of the
mode eorts Eff modes .
The concept enables to correctly turn the design
key by respecting the most critical mode and the
location [25,29] in the Finite Element idealization
of the structure (Appendix A3).
Homogenisation of the UD-material comes to its
limit if a constituent stress governs the failure.

Fig. 30. (a) Stressstrain curves ^ y :^ x =1:1 [+45/45]s-laminate. E-

glass/MY750 [3] (without temperature drop). No.14 of Table 11 [3].
(b) Stressstrain curves ^ y :^ x =1:1 [+45/45]s-laminate. E-glass/
MY750 [3] (with temperature drop 12022 , as desired in addition by

Fig. 32. Stressstrain curves ^ y :^ x =1:1 [+45/45]s-laminate. E-

glass/MY750 [3]. No. 13 of Table 11 [3].

Fig. 31. Stressstrain curves ^ y :^ x =0:1 [0/90]s-laminate. E-glass/ Fig. 33. Stressstrain curves ^ y :^ x =1:0. [0/+45/45/90]s-laminate.
MY750 [3] No. 12 of Table 11 [3]. AS4/3501-6 epoxy [3] No. 7 of Table 11 [3].
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 367

Fig. 34. ( 2 , 21)-IFF curves, tubes [6], GFRP: E-glass/LY556/HT976/DY070, CFRP: T300/LY556/HT976.

This is the case for Fjj , where the macro- laminae and fabric laminae. For other textile
mechanical stress  1 has to be replaced by the preforms (3D, stitched etc.) engineering models
actual bre stress  1f. A bre stress may be zero have to be developed.
not even for zero  1. Therefore,  1f has to be The transferability to rhombically-orthotropic
estimated as  1f=e1 . E1f. In order to remain on composites and other materials (fabrics) works
composite stress visualization level  1f will be [27,28].
multiplied by the bre volume fraction vf. The choice of linear or other terms of stress
The mode t avoids the shortcomings of the invariants is based on whether there are volume
global t which maps the course of test data by and/or shape changes of the material element as
mathematically linking failure modes which are well as on curve tting considerations.
in reality not mechanically linked. One typical In respect of the scatter of the actual test data the
shortcoming is, that a reduction of the strength parameter set b? 1; b?jj b?jj 0 will often be
of one mode might increase the multi-axial an approach good enough for nal failure ana-
strength in another (independent) mode or part lysis of the laminate.
of the global failure surface.
For the prediction of nal failure the initial fail-
ure approach is not of that high concern, if wedge Acknowledgements
failure, caused by F? < 1 and followed by dela-
mination failure, will not occur (see Pucks drive The authors gratefully express their thanks for the
shaft [42] or torsion spring). intensive collaboration with Professor A. Puck review-
Each failure condition describes the interaction ing this paper in the context with the comparison of the
of stresses aecting the same failure mode and two approaches. The authors also thank the reviewers
assesses the actual state of stress in a material point. for their constructive comments.
For ( 2, 3) states of stress Mohrs stresses,
Mohrs envelope curve, and the inclined fracture
angle fp may be determined. Appendix A
Damage mechanics is captured in the FMC
conditions so far as the stiness reduction is A.1. Comparison of Pucks and Cuntzes failure theories
determinable via the ( eq,e)-curve, and by the
predictability of delamination initiation, apply- A.1.1. Comparison of Pucks fracture plane based IFF-
ing F? and F? . criteria and Cuntzes FMC-based invariant formulations
Regarding the investigations in theory and test The two Sections A1.1 and A1.2 are a common for-
carried out in Germany on the lamina material mulation of Puck and Cuntze, because both authors
level in the last years (still going on) the under- have often been asked for an explanation of the coin-
standing has improved a lot and seems to be a cidences and dierences between their approaches. The
good basis to tackle laminates stacked-up of UD- following should be mentioned in this context:
368 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

With respect to the dierent eort that has been Pauls modication of the CoulombMohr theory of
put by Puck et al. (incl. Cuntze) into the fracture fracture [39] is valid. This is based on the assumption,
plane based criteria and by Cuntze into the FMC that two dierent modes of fracture can occur which
based criteria the Puck criteria are approaching leads to the following fracture hypothesis (formulation
the series status and the Cuntze criteria only analogue to that for isotropic material): An intrinsically
the development status. brittle material will fracture in either that plane where
The FMC criteria seem to be generally applicable the shear stress  nt reaches a critical value which is given
to all materials. Therefore, there are a few short by the shear fracture resistance R?? of a bre parallel
comings in their application to UD-material. plane increased by a certain amount of friction. This
friction is caused by the simultaneously acting com-
As early as 1968/1969 Puck concluded from pressive stress  n on that plane. Or, it will fracture in
experimental observations that two completely dierent that plane, where the maximum principal stress ( II or
types of fracture should be distinguished and  III) reaches the transverse tensile strength R?t (Fig. A1/
theoretically treated by separate failure criteria: Fibre 1 and A1/2). For reasons of simplicity the bars over R
Failure (FF) and Interbre Failure (IFF) [8,9]. In the are skipped in Chaper A1.
early seventies the discrimination of these two fracture
types became common practice in the German A1.1.2. Results for plane stress ( II,  III)
aerospace industry[46]. In all later papers of Puck and For this state of stress without any longitudinal shear
Cuntze the separate treatment of FF and IFF has ( 31, 21) there is a complete coincidence of the formula-
been maintained. For FF both authors use simple max- tions of Puck and Cuntze.
imum stress criteria, based on the consideration, that The treatment of this problem by Mohrs circle (repre-
the composite fails when the bres reach a certain cri- senting the state of stress ( n, nt), on any plane, see
tical stress. Fig. A1/1 and Mohrs fracture envelope [representing the
Both authors feel that for the new anisotropic bres a fracture limit for combined ( n, nt)-stresses] is well known.
better approach for FF prediction may be necessary.
Since another fundamental paper of 1992 [12], For the domain  n < 0 Puck [42] starts with the
research in Germany has concentrated on the improve- assumption of a parabolic fracture envelope
ment of IFF criteria. This appeared to be of higher  nt= nt( n), that reads:
importance than assumed in the past after it had been  A 2
learned from experience on torsion tube springs that the
nt R?? 2p??
R?? n A1
wedge eect of oblique fractures under transverse com-
pression can cause destruction of the whole composite wherein R?? is the transverse shear fracture resistance
part [47]. Besides this, under alternating loads, micro- of a bre parallel plane against its fracture caused by a
cracks, due to IFF (caused mainly by transverse tensile -stressing acting on that plane. The coecient p?? 
is a
stress), give rise to high peaks of interlaminar stresses so-called friction coecient for  n < 0.
which initiate local delaminations. At fracture Mohrs circle and the fracture envelope
have a common point of contact, that means the same
A.1.1.1. Common foundation of the two approaches. The
failure theories of Puck and Cuntze are based on the
same fundamental assumptions:

The UD-layer is transversally-isotropic and fail-

ure occurs by brittle fracture.
Mohrs statement is valid: The material strengths
are determined by the stresses on the fracture
The fracture plane may be inclined with respect
to the plane which the external stresses are acting
on. This is, for instance, true for uni-axial
transverse compression.
For states of stress without longitudinal shear
( 31, 21), that means plane stress conditions Fig. A1/1. Mohrs fracture envelope and some Mohr circles for frac-
turing stresses (a) for uniaxial transverse compression  II, (b) limiting
consisting of a stress state ( 2, 3, 23) which can
circle for simultaneous shear fracture (SF) and cleavage fracture (CF)
be replaced by ( II, III), the so-called principal on dierent action planes. Between SF and CF no circle can touch the
stresses of the transversally-isotropic plane, both fracture envelope, (c) for pure shear  23= II= III, (d) for uniaxial
authors make the same assumption: transverse tension  III.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 369

II R?t or III R?t ;

see Fig. A5 or A1/2.

A1.3. Results for states of stress with additional

longitudinal shear (t31,t21)

In this eld the two authors use rather dierent


Puck stays with the physically based con-

sideration of the mechanical interaction of the
stresses  n, nt, n1 on the fracture plane (Fig. A1/
3). He uses simple polynomials (parabolic or
elliptic) to formulate a master-fracture body in
the ( n, nt, n1)-stress space.

Starting from this (master-) fracture body generally

Fig. A1/2. Fracture curve ( II,  III ) resulting from Fig. A1/1 with
no analytical solutions can be found for the fracture
tensile cut-os and typical fracture angles /fp for uniaxial transverse angle /fp (with the exception of ( 1, 2, 21)-states of
tension and compression. stress) and therefore no analytical solutions can be
given for the fracture bodies in  1, 2, 3, 23, 31, 21.
Therefore, the necessary search for the fracture
inclination d nt/d n. From this condition the fracture plane, that means for the plane with the lowest
angle /fp between the action plane of  II and the frac- reserve factor minfRes(/) or the highest stress expo-
ture plane can be calculated, which is varying a little sure factor maxfE(/), has to be done numerically
with the dierence of ( II III), (using the formulation of the fracture condition in
R?c  n, nt, n1) in an angle range 90 4/ 4+90 . By
cos2/fp cos2/cfp for n < 0 A2 means of the found fracture angle fp, resulting from the
numerical procedure, the stresses 1 ; 2 ; :::21 at fac-
with /cfp =fracture angle under uniaxial transverse ture can be calculated by multiplying the acting stresses
compression (angle between the action plane of the ( 1, 2,  21) by the lowest reserve factor minfRes(/)=
uniaxial compressive stress  II and the corresponding fRes(/fp).
fracture plane which is 45 < j/cfp j < 60 ) and The numerical search for the fracture plane is an
R?c =transverse compression strength. In this equation inconvenience, but on the other hand the user of this
 II and  III are stresses at fracture! approach automatically gets an information on the
By means of this result a denite form in  II, III for fracture angle and on the fracture mode. Puck denes
the fracture condition is found which is parabolic and the fracture mode as the stress combination ( n, nt, n1)
invariant in the transversal plane: or (,,?||) on the fracture plane. For the calculation of
a? b? the fracture stresses Cuntzes invariant formulation is of
F? c II III  c 2 II  III 1: A3 course the more convenient one.
R? R ? The results can be visualized by fracture bodies in a 3-
dimensional ( II, III, o1)-space, where  o1 is the resul-
Cuntze in contrast to Puck starts already with tant of  31 and  21. These fracture bodies are not sym-
this invariant formulation (A3): The adaptation metric with respect to the ( II= III)-plane [21].
to experimental uniaxial compression results
(strength R?c and fracture angle /cfp ) gives Cuntze uses three simple invariant formulations
  in ( 1, 2, 21)-one linear, one quadratic and one
a? b?  1 and b? 1 2cos2/cfp 1 cubic polynomial- which lead to fracture bodies
in the ( II, III, o1) space similar to those of Puck.
In Eq. (21b) another adaption of b? to test results is He feels that mechanical and probabilistic inter-
shown. Pucks and Cuntzes approach for the domain actions cannot be clearly distinguished and
 n < 0 are connected by the relation (A2) for the frac- therefore he models the mode interactions by a
ture angle /fp. For the domain  n 50 both authors use simple probabilistic series model (rounding-o
the tensile cut-os recommended by Paul [48]. That procedure achieved by the determination of f Res
means that the fracture stress is either or E(res)).
370 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

Fig. A1/3. Action planes and fracture causing Mohr stresses.

Attention has to be paid to the fact that the Of course, one has to pay for the higher convenience
expression mode has dierent meanings in the of the invariant approach with a certain loss of physical
papers of Puck and Cuntze. Puck dierentiates correctness and the inability to predict the fracture
between seven interbre fracture (sub)-modes angle for states of stress including longitudinal shear
M1M7 (according to the number of the possible  21, 31, However, this may be acceptable in many cases
stress combinations acting on the fracture plane) of design practice.
which may be allocated to the three Modes A, B,
C (see Fig. A1/4): A1.4. Comparison of Pucks and Cuntzes failure
analysis of laminates
Group with  n 50
  This section focuses on a 2D-laminate failure analysis
M1 ?t ; ?? ; ?jj the mostgeneralmode as performed in the failure exercise, Part A[3].
M2 ?t ; 0; 0 >
 t =
M3 ? ; 0; ?jj Domain ofModeA A5
  > ;
M4 0; 0; ?jj

Group with  n < 0

M5 ?c ; 0; ?jj Domain ofModeB

M6 ?c ; ?? ; ?jj

 Domain ofModeC
M7 ?c ; ?? ; 0

Cuntze uses the expression mode to address his three

dierent invariant IFF conditions, based on the idea that
for each of these fracture conditions in their pure regimes
either the ?t -, the ?c , or the ?k -stressing is dominant. Fig. A1/4. ( 2, 21)-fracture curve with IFF-modes A, B, C.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 371

For bre failure (FF) of the UD-lamina both authors Mode A is valid for  2 50 and combines the
use the same simple maximum stress failure criterion: modes M2, M3, M4 mentioned in Eq. (A5)
A c
1 Mode B is valid for 04| 2/ 21|4R?? /|21 | and is
f EFF t 1 for 1 5 0 identical with mode M5.
1 Mode C is valid for the region with fp 6 0, i.e.
and f EFF c 1 for 1 < 0: A6 c
04| 21/ 2|4|21 A
| /R?? and combines the modes
M6 and M7.
fE is the stress exposure factor used by Puck. It has
essentially the same meaning as Cuntzes resultant stress One should remember that the expression mode has
eort E(res). The value of fE or E(res), respectively, dierent meanings in the papers of Puck and Cuntze!
quanties the risk of fracture. Fracture occurs, if Pucks stress exposure factors f EIFF for his Modes A, B
fE=1=100%. and C are not equivalent to Cuntzes mode eorts
Both authors also assume that FF in at least one E(mode) but to Cuntzes E(res)! Like f EIFF also E(res)
lamina of a laminate means nal failure of the laminate. quanties the risk of fracture due to the combined
Therefore, the biaxial failure envelopes for nal fail- action of  2 and  21.
ure of laminates predicted by the two authors do not Pucks fracture condition for IFF of a UD-lamina is
dier very much, as long as the laminates have three or
f EIFF 1: A8
more bre directions. The strengths of these laminates
are bre dominated.
Also, the predicted stress/strain curves of such lami- For a UD-lamina in a laminate, this means the onset
nates look very similar because the bres which are of progressive IFF (matrix cracking) the three dierent
much stier than the matrix carry the main portion of equations for f EIFF are (Fig. A1/4):
the loads. Dierent degradation procedures after the
onset of interbre failure (IFF) do therefore not inu- For Mode A:
ence the predicted strains very much. This is especially 1
true for CFRP laminates. f EIFF
Pucks degradation procedure (known as the -degra- 2s 3
dation) for the secant moduli E2(sec) and G21(sec) after # $2
2 R?jj  p
 4 21 5
22 p?jj 2 ; A9
the onset of IFF is rather simple, since Pucks IFF- R?t ?jj
criteria are completely based on the assumption of a
mechanistic interaction of  2 and  21. Probabilistic
aspects can be dealt with, if necessary, in a separate
operation [42]. For Mode B:
In Pucks theory the numerical search for the fracture #q $
IFF 1 2  2 
angle fp, that means the search for the stress action fE  21 p?jj 2 p?jj 2 ; A10
plane with the highest angle dependent stress exposure
factor max f EIFF ( ), is not necessary in the special case
of a plane state of stress ( 1, 2, 21). For tensile  2 > 0 For Mode C:
and also for moderate compressive stress j2 j < 0:4R?c
R?c 2
21 2
the fracture plane is the same as the action plane of  2 f EIFF  2  : A11
2 R?c
and  21 (Fig. A1/4). That means: fp=0. For rather high 4 R?jj p?jj
compressive stresses j2 j > 0:4R?c the fracture angle
pf 6 0 can be calculated from a very simple analytical
expression: Hence, for Mode A the fracture angle fp is 0 and,
s because  2 is a tensile stress, the micro cracks tend to
R?? open. The resulting decrease of the secant moduli E2(sec)
fp arctan : A7
2 and G21(sec) is modelled by Puck by a simultaneously
starting degradation of E2(sec) and G21(sec). That means,
Attention! In this equation  2 is the compressive stress secant moduli E2(sec) and G21(sec) with  < 1 are used
at fracture caused by a combined ( 2,  21) state of stress. after the onset of IFF. The degradation factor  is a
Based on the knowledge of the fracture angle there decaying function decreasing with increasing load, in
have been found three simple analytical expressions for order to keep f EIFF =1. After the onset of IFF only
the stress exposure factor f EIFF formulated with  2 and average stresses can be calculated for a micro-cracked
 21 instead of  n, nt, n1. Each of the three equations is lamina. Average stresses are dened as stresses smeared
valid for a certain region of the ( 2, 21)-fracture curve) over some length of the cracked lamina (which includes
[42,47]: a number of micro-cracks).
372 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

After the onset of IFF, Puck calculates the average failure modes the secant moduli E2(sec) and G21(sec) are
stresses  2() and  21() by using .E2(sec), mG21(sec), taken from the  2(2)-curve or the  21( 21)-curve not just
and .12. He assumes that in the progressive cracking at the stresses  2 or  21 resulting from the stress and
process of a lamina its average stresses  2() and  21() strain analysis for the actual level load. Their values are
remain approximately constant with increasing load. taken at a little higher stress in the hardening branch
This is achieved in the calculation by keeping with increasing stress and at a little lower stress in the
f EIFF =1=constant after the rst IFF has occurred. softening branch with decreasing stress. This stress
In contrast to Pucks completely f EIFF -controlled - correction is controlled by the so-called triggering
degradation Cuntze uses his stress exposure factor approach, which is described in Section 6 [see Eqs. (35)
E(res) (or fE) only for a certain correction, in order to and (36)]. The controlling parameter is the ratio of the
take into account mode-interactions. His degradation is resultant stress exposure factor E(res) to the maximum
mainly controlled by using the  2("2)- or the  21( 21)- mode exposure factor maxE(mode). By this triggering
stress/strain curve for nding the valid secant modulus approach lower secant moduli E2(sec) and G21(sec) are
E2(sec) or G21(sec), respectively. The branches with provided for the next calculation loop as those which
increasing stresses (hardening) of these stress/strain would result without the correction by the triggering
curves are found by the usual experiments with uni-axial approach.
 2-stress or pure  21-stress, respectively. The branches Figs. A1/5 and A1/6 visualize Pucks -degradation
with rapidly decreasing stresses (called softening) are and Cuntzes triggering approach.
preliminarily assumed (see Figs. 1017). Like Puck, In Cuntzes theory for the actual load the degradation
Cuntze calculates the stresses  2 and  21 in the laminae of E2(sec) and G21(sec) is performed with the same trigger
of the laminate by using secant moduli from the  2("2)- factor TrF. In contrast to Pucks theory, if one of the
and  21( 21)-stress/strain curves. However, Cuntze has corrected equivalent mode stresses has reached its
to pay attention to a proper interaction of the inter- strength level, a rapid decrease of the modes average
active modes in the stress and strain analysis in the fol- (smeared over the micro-cracks) equivalent stress will
lowing manner: In order to take into account the follow. There is another dierence: Cuntzes triggering
combined probabilistic/mechanistic interaction of the approach is already active before the onset of IFF. This

Fig. A1/5. Schematic illustration of Pucks assumptions about the

average stresses 2  and 21  after the onset of IFF. The strain "repr: Fig. A1/6. Schematic illustration of Cuntzes assumptions about the
is a representative strain, which is proportional to the load on the stresses  2 and  21 before and after IFF-initiation. Results of the
laminate. triggering approach.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 373

can perhaps be justied by the fact that there is a certain A.3. Visualization of the reserve factors of a uniformly
mutual interaction of  2 and  21 on their strains before loaded sandwich plate
the fracture stresses have been reached [see Fig. 1] in the
literature [47]. In this Appendix a still existing example (Table A3/1)
Due to the severe lack of experimental experience is taken in order to visualize the handling with the
mode res
about the real degradation of laminates after IFF initia- values computed for f Res and f Res (further see
tion dierent authors make very divergent assumptions Table A3/2).
about the average stresses in a lamina after the onset of Failure conditions and corresponding curve para-
IFF until nal failure of the laminate, as can be seen meters are not depicted here, because they belong to a
from the failure exercise [4]. Figs. A1/5 and A1/6 slightly dierent former set of fracture conditions [26].
demonstrate this for the two authors Puck and Cuntze. But, the following results nevertheless highlight how the
In order to demonstrate the dierent assumptions designer will work with reserve factors or with eorts.
most drastically the case is considered where compres- The listing of the reserve factors values in Table A3/2
sive  2 does not lead to a premature nal failure due to completely describes the stress situation in all the lami-
the wedge eect. Because of the assumption that for nae, with (k) indicating the particular lamina. Numbers
compressive  2 the cracks stay closed after IFF initia- in quadratic brackets refer to the nite element (the FE
tion Puck does not degrade E2s and 12. Therefore a code MARC was employed).
rapid -degradation of  21 alone has to keep f EIFF =1. Lamina k=3 contains the minimum resultant reserve
In contrast to this Cuntze treats the compressive stress factor minf Res (0.82), which is due to the 90 -angle
 2 similar to a tensile stress  2 with a pronounced soft- still a little smaller than that for the 45 -lamina (k=2)
ening branch of the stress/strain diagram (Fig. 11). He is for which Fig. A3/1 is prepared. This plot depicts the
basing this approach on weakening eects due to the distribution of the resultant reserve factor for the
cracking. lamina ( =+45 , k=2) of the sandwich plate. Its
smallest value 0.89 is also lower than 1, thus indicating
A.2. Additional biaxial (s2, t21)-test data for IFF which would cause a redesign if IFF is not per-
UD-laminae mitted. The IFF is caused by ?t in the laminas plane
and is critical over a large domain of the lamina. Similar
Fig. 34 provides additional test data for one GFRP to fRes Table A/2 can be lled in by the various E. A
and one CFRP material [6]. As test specimen the wound clear determination of the design driving maxE(mode)
tension/compression- torsion tube specimen was used. would then be pointed out, too.

Table A3/1
Geometry and loads of the GFRP sandwich plate. Mechanical properties (mean values) and strength properties (f=0.60) and data of calibration
points (mean values)
374 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

A.4. Simplications of FMC-based failure conditions  

1 2c 3c =R1 c?
b? q
 c   2 :
  c =R1 c  c   c =R1 c2
2 3 ? 2 3 ?
Simplication of F? : As still briey mentioned,
the F? : condition may be homogenized (I4 is not
homogeneous to I2) in the form
This value is dierent to that of the former bc? ,
p of course, because F? is a new one.
   I2 I4
F3 : F? b?  1 
b? 1; A12 Now, the reserve factor is linearly computable
R? c 1
Rc? due to
f ?
Res R? = b?  1 I2 b? I4 : A14
which means a replacing of I4 by I4 and of
setting b?jj 0: Then, the curve parameter b? Determination of max I3/2
3 for the case dis-
may be simply determined from the equation crimination:

Table A3/2
Reserve factors calculated for all laminae [12]
fRes k fjj
Resk fjj
Resk f?
Resk f?
Resk fjjResk

k=1 0 1.33 [142] 12.8 [202] 100. [..1] 1.34 [142] 3.13 [141] 3.51 [142]
k=2 45 0.87 [267] 27.6 [349] 53.3 [340] 0.893 [267] 1.75 [267] 2.26 [267]
k=3 90 0.82 [686] 38.6 [574] 41.2 [606] 0.820 [686] 4.62 [559] 2.15 [686]
k=4 45 0.94 [750] 31.6 [821] 87.0 [762] 0.964 [750] 1.89 [766] 2.44 [750]
k=5 0 1.92 [1146] 14.9 [996] 100. [961] 1.94 [1146] 5.21 [1146] 5.25 [1146]
k=6 core Not relevant here
k=7 0 4.02 [1618] 100 [1441] 11.9 [1567] 4.11 [1618] 6.36 [1667] 6.84 [1618
k=8 45 1.39 [1880] 100 [1681] 25.3 [1681] 1.47 [1880] 2.27 [1880] 2.90 [1880]
k=9 90 2.99 [1984] 51.4 [2054] 30.9 [2029] 3.52 [2069] 5.25 [2029] 2.99 [1984]
k=10 45 1.29 [2336] 66.6 [2214] 22.1 [2243] 1.37 [2336] 2.11 [2298] 2.68 [2336]
k=11 0 2.27 [2583] 100 [2401] 10.2 [2640] 2.31 [2583] 3.55 [2583] 4.56 [2582]

Fig. A3/1. Reserve factor f Res of failure responsible lamina k=2 (lower skin) [G. Sukarie, MARC][25].
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 375

Fig. A5. The ( 21,  2,  3 )- failure body.

The limit for the applicability of F?k for the A.5. The (t21,s2,s3 )- failure body
given state of stress (marked by a dot) is
: : : : The most interesting partial IFF body is that for the
3 =I 2 I 3  I 5 4  b?jj ; stress combination ( 21, 2, 3). By this failure body (a
dicult and time consuming work to produce it) the
with a preliminary to be conrmed w=1.1. main dierences of the new IFF conditions are dis-
From the ratio above the limiting maximum played: This is at rst the dierence between ( 21, 2)
value on the failure surface can be deduced via and ( 21, 3), see the typical asymmetry outlined in var-
ious papers of Puck et al. According to the rounding
I33=2 b?jj I2 I3  I5 R?jj
; procedure the 21 ; 3t R?t front side is not fully ver-
tical anymore as by Puck documented. Nevertheless it
R?jj provides with the  main information. Secondly the
I33=2 =I2 I3  I5 b?jj  1 ; rounding-o in the 2t ; 3t -domain is depicted.
I2 I3  I5 The post-processing of the failure body has caused
3 some smaller irregularities and should be reworked with
 b?jj b?jj R?jj =maxI2 I3  I5 : A15
a better tool.

From max(I2I3I5) follows

maxI33=2 R?k
- b?k maxI2 I3 -I5 : A16
[1] Hinton MJ, Soden PD, Kaddour AS. Comparison of failure
prediction methods for glass/epoxy and carbon/epoxy laminates
under biaxial stress. ICCM11 1997;V:67282.
Further simplication of f?11 (recommendation): [2] Hinton MJ, Soden PD, Kaddour AS. Failure criteria in bre-
reinforced-polymer composites (special issue). Composites
 p p Science and Technology 1998:58.
F?k 2 I3 b?jj 3 I2 I3  I5 =R?jj [3] Soden PD, Hinton MJ, Kaddour AS. Lamina properties, lay-up
 p p A17 congurations and loading conditions for a range of bre-rein-
f Res R?jj = 2 I3 b?jj 3 I2 I3  I5 : forced composite laminates (special issue). Composite Science
and Technology 1998;58:101122.
376 R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377

[4] (a) Soden PD, Hinton MJ, Kaddour AS. A comparison of the terion in comparison with common strength criteria. Proceedings
predictive capabilities of current failure theories for composite of ICCM-12, Paris, France, July 1999.
laminates (special issue). Composites Science and Technology [22] Hufenbach W, Kroll L. A new failure criterion based on the
1998;58:122554. mechanics of 3-dimensional composite materials. ICCM-10,
(b) Kachanov, L.M. Foundations of the Theory of Plasticity. Whistler, Canada, 1995.
North-Holland (1971). [23] Cuntze RG. Fracture-type strength criteria formulated by invar-
(c) Betten J. Recent advances in mathematical modelling of iants which consider the materials symmetries of the isotropic/
material behaviour. Math Comput Model 1990;14:3751. anisotropic material used. Conf. on Spacecraft Structures, Mate-
[5] Knappe W, Schneider W. Bruchkriterien fur unidirektionalen rials and Mechanical Testing. [ESA-CNES-DARA]. Noordwijk,
Glasfaser-Kunststo unter ebener Kurzzeit- und Langzeitbean- March 1996 [Conf. Handbk].
spruchung. Kunststoe, Bd 1972;62:8648. 23a VDI2014: German guideline on development of FRP com-
[6] VDI97 Cuntze RG, et al. Neue Bruchkriterien und Festigkeits- ponents analysis; 2003 (in German and English).
nachweise fur unidirektionalen Faser-kunststoverbund unter [24] Cuntze RG. Evaluation of multi-axial test data of UD-laminae
mehrachsiger BeanspruchungModellbildung und Experimente. by so-called fracture type strength criteria and by supporting
VDI-Fort-schrittbericht, Reihe 1997;5(506):250. probabilistic means. ICCM-11, Gold Coast, Australia, 1997.
[7] Herrmann G. Zum Bruchverhalten gerichteter Glasfa- [25] Cuntze RG, Sukarie G. Eective dimensioning of 3D-stressed
serverbunde. Dissertation, TU-Stuttgart, Inst. f. Werkstoe im UD-laminae on basis of fracture-type strength criteria. Con-
Bauwesen, 1982. ference handbook, Presentation Int. conf. on Mechanics of
[8] Puck A. Calculating the strength of glass bre/plastic laminates Composite Materials. Riga, 2023 April 1998.
under combined load. Kunststoe, German Plastics [German [26] Cuntze RG. The failure mode concepta new comprehensive
Plastics is the bilingual English and German edition of Kunst- 3D-strength analysis concept for any brittle and ductile behaving
stoe] 1969;55:1819 [in German. p. 7807]. material. Europ. Conf. on Spacecraft Structures, Materials and
[9] Puck A, Schneider W. On failure mechanisms and failure criteria Mechanical Testing. [ESA-CNES-DGLR-DLR] Braunschweig,
of lament-wound glass-bre/resin composites. Plast Polym 1969: November 1998 [ESA SP-428]. p. 26987.
3343. [27] Cuntze RG. Strength prediction for multi-axially loaded CMC-
[10] Puck A. Praxisgerechte Bruchkriterien for hochbeanspruchte materials. In: 3rd European Workshop on thermal Protection
Faser-Kunststoverbunde. Kunststoe 1992;82(2S):14955 Systems. [ESA-ESTEC] Noordwijk, March 1998 [WP P141].
[Fracture criteria for highly stressed bre plastics composites [28] Cuntze RG. Application of 3D-strength criteria, based on the so-
which meet requirements of design practice. Kunststoe German called failure mode concept, to multiaxial test data of sandwich
Plastics 1992;82(2):368]. foam, concrete, epoxide, CFRP-UD lamina, CMC-fabric
[11] Puck A. Faser-Kunststo-Verbunde mit Dehnungs- oder Span- Lamina. ICCE/5, Las Vegas, July 1998 [presentation].
nungs-Kriterien auslegen? Kunststoe 1992;82(5):4314 [Should [29] Cuntze RG. Progressive failure of 3D-stressed laminates: multiple
bre-plastics composites be designed with strain or stress criteria? non-linearity treated by the failure mode concept (FMC). DUR-
Kunststoe German Plastics 1992;5:3436]. ACOSYS 99, Brussels, July 1999.
[12] Puck A. Ein Bruchkriterium gibt die Richtung an. Kunststoe [30] Boehler JP. Failure criteria for glass-ber reinforced composites
1992;82(7):60710 [A failure criterion shows the direction under conning pressure. J Struct Mechanics 1985;13:371.
further thoughts on the design of laminates. Kunststoe German [31] Boehler JP. Personal note to the rst author on fabric invariants;
Plastics 1992;82(7):2932]. 1995.
[13] Hashin Z. Failure criteria for unidirectional bre composites. J of [32] Jeltsch-Fricker R, Meckbach S. A parabolic Mohr fracture con-
Appl Mech 1980;47:32934. dition in invariant formulation for brittle isotropic materials.
[14] Cuntze RG. A physically based 2D/3D-inter-bre-failure criter- ZAMM 1999;79:46571 [in German].
ion for brittle UD-layersHashins idea and Pucks realization. [33] Meckbach S. Invariants of cloth-reinforced bre reinforced plas-
Key-note lecture, ISODUR 93, Porto, July 1821. tics. Kasseler Schriften zur angewandten Mathematik 1998;1 [in
[15] Jeltsch-Fricker R. Bruchbedingungen vom Mohrschen Typ fur German].
transversal-isotrope Werkstoe am Beispiel der Faser-Kunststo- [34] Christensen RM. The numbers of elastic properties and failure
Verbunde. ZAMM 1996;76:50520. parameters for ber composites. Transactions of the ASME 1998;
[16] Jeltsch-Fricker R, Meckbach S. Fast solver of a fracture condi- 120:1103.
tion according to Mohr for unidirectional bre-polymer compo- [35] Cuntze RG. Deterministic and probabilistic prediction of the
site. Scripts of the University of Kassel on Appl. Math., No. 1/96 distribution of inter-bre failure test data of prestrained CFRP
[in German]. tubes composed of thin layers and loaded by radial pressure.
[17] Michaeli W, Huybrechts D. A new approach for the dimension- Wollongong. In: Advanced Composites 93. The Minerals,
ing of thick laminates using physically based strength criteria. In: Metals & Materials Society, 1993. p. 579585.
Proc. 39th Conf. of the Society for the Advancement of Material [36] Grimmelt M, Cuntze RG. Probabilistic prediction of structural
and Process Engineering (SAMPE), vol. 2, 1114 April 1994, test results as a tool for the performance estimation in compo-
Anaheim, CA p. 282940. site structures design. Beuth Verlag, VDI-Bericht 1989;771:191
[18] Huybrechts D. Ein erster Beitrag zur Verikation des wirkebe- 200.
nebezogenen Zwischenfaserbruchkriteriums nach Puck. Disserta- [37] Rackwitz R, Cuntze RG. System reliability aspects in composite
tion an der RWTH Aachen; 1996. structures. Eng Opt 1987;11:6976.
[19] Rackwitz R, Gollwitzer S. A new model for inter-bre-failure of [38] Cuntze RG. Failure path analysis of multi-layered bre rein-
high strength uni-directionally reinforced plastics and its relia- forced plastic components with the reliability calculation pro-
bility implications. NATO-workshop PROBAMAT-21 Century, gramme FRPREL. Noordwijk, October 1987.
Perm, Russia, 1012 September 1997. [39] Mohr O. Welche Umstande bedingen die Elastizitatsgrenze und
[20] Kopp J, Michaeli W. Dimensioning of thick laminates using new den Bruch eines Materials? Civilingenieur 1900;XXXXIV:1524
IFF strength criteria and some experiments for their verication. 30, 15727.
In: Proceedings Conf. on Spacecraft Structures Materials and [40] Flaggs DL, Kural MH. Experimental determination of the in situ
Mechanical Testing, ESA, 2729 March 1996. transverse lamina strength in graphite epoxy laminates. J Comp
[21] Kopp J, Michaeli W. The new action plane related strength cri- Mat 1982;16:10316.
R.G. Cuntze, A. Freund / Composites Science and Technology 64 (2004) 343377 377

[41] (a) Masters J. Fractography of modern engineering materials. [52] Goldenblat II, Kopnov VA. Strength of glass-reinforced plastics
Composites and metals, 2nd vol. ASTM STP1203; 1994. in the complex stress state. Polymer Mechanics of Mechanical
(b) Ashton JE, Whitney JM. Theory of laminated plates. Progress in Polimerov 1966;1:549.
Materials Sciences Series, vol. I. Stamford, Technomic Publ. 1970. [53] Thom H. A review of the biaxial strength of bre-reinforced
[42] (a) Puck A. Festigkeitsanalyse von Faser-Matrix-Laminaten plastics. Composites Part A 1998;29:86986.
Modelle fur die Praxis. Munchen: Carl Hanser Verlag; 1996. [54] Tsai SW, Wu EM. A general theory of strength for anisotropic
(b) Lekhnitskii SG. Theory of elasticity of an anisotropic elastic materials. J Comp Mater 1971;5:5880.
body. San Francisco: Holden Day; 1963. [55] Yeh HY, Kilfoy LT. A simple comparison of macroscopic failure
[43] Awaji H, Sato S. A statistical theory for the fracture of brittle criteria for advanced ber reinforced composites. J Reinforced
solids under multi-axial stresses. Int J Frac 1978;14:R1316. Plastics Comp 1998;17:40645.
[44] MiL hdbk 17. Plastics for aerospace vehicles, vol. I, reinforced plas- [56] Wang JZ. Failure strength and mechanism of composite lami-
tics; vol. II; vol. III utilization of data. USA: Dep. of Defence (DOD). nates under multiaxial loading conditions. Dissertation, Univ. of
[45] Rolfes R, Noor AH, Rohwer K. Ecent calculation of transverse Illinois at Urbana, 1993.
stresses in composite plates. MSC-NASTRAN User Conference, [57] Rowlands RE. Strength (failure) theories and their experimental
1997. correlation. In: Sih GC, Skudra AM, editors. Handbook of
[46] (ZTL80) Dornier, Fokker, MBB, DLR. Investigations of fracture composites, band III, Chapter 2. Madison (WI, USA): Elsevier
criteria of laminae. Grant from German ministry, BMVg [multi- Science; 1985. p. 71125.
axial testing] 19751980 reports in German. [58] Matzenmuller A, Lubliner J, Taylor RL. A constitutive model for
[47] Puck A, Schurmann H. Failure analysis of FRP laminates by anisotropic damage in ber composites. Mechanics of Materials
means of physically based phenomenological models [special 1995;20:12552.
issue]. Comp Sci Technol 1998;58. [59] Sukarie G. Einsatz der FE-Methode bei der Simulation des
[48] Paul B. A modication of the Coulomb-Mohr theory of fracture. progressiven Schichtversagens in laminierten Faserverbund-
J Appl Mech 1961:25968. strukturen. Symposium Berechnung von Faserverbundstrukturen
[49] Christensen RM. Stress based yield/failure criteria for ber com- unter Anwendung numerischer Verfahren. Munchen, Techn.
posites. Int J Solids Struct 1997;34(5):52943. Univ., 13/14 March 1996.
[50] Christensen RM. Yield functions/failure criteria for isotropic [60] Slight DW, Knight NF, Wang JT. Evaluation of a Progressive
materials. Proc R Soc Lond A 1997;453:147391. failure analysis methodology for laminated composite structures.
[51] Hart-Smith LJ. An inherent fallacy in composite interaction 38th Structure, Structure Dynamic and Material Conference,
failure curves. Designers Corner, Composites 1993;24:5234. April 1997 [AIAA Paper 97-1187].